Information processing device, information processing method, and program

Abstract

[Problem] Conventionally, it has not been possible to simulate brain activity when performing an operation with consideration for long-term memory and short-term memory. [Solution] The problem can be solved by an information processing device 1 comprising: a reception unit 12 that receives information; a feature acquisition unit 131 that acquires a feature set from the information; a transfer unit 132 that performs a transfer process for determining one or more node identifiers paired with a start point condition which matches the feature set and determining a firing node identifier of a node to be fired, which is connected by an edge to a fired node that is identified by each of one or more node identifiers; and a corpus callosum unit 133 that acquires a long-term firing pattern and a short-term firing pattern from one or more firing node identifiers which have been determined by the transfer unit 132, determines, from among an operation paired with a firing pattern that is similar to the long-term firing pattern and an operation paired with a firing pattern that is similar to the short-term firing pattern, an operation which matches an operation condition, and performs the determined operation.

Description

Information processing device, information processing method, and program

[0001] The present invention relates to an information processing device or the like that simulates processing in the brain.

[0002] Conventionally, there has been an augmented reality system that claims to be able to monitor an individual's brain state (see Patent Document 1).

[0003] Special table 2021-511612 publication

[0004] However, conventional technologies have not been able to simulate the behavior of the brain when performing actions that take into account long-term memory and short-term memory. More specifically, when actions stored in long-term memory are often performed, adult-like behavior is obtained, and when actions stored in short-term memory are often performed, childish behavior is obtained. However, conventional technologies have not been able to simulate adult-like or childish behavior. Note that adult-like behavior can be said to be behavior based on calm judgment or behavior based on past examples, etc. Childish behavior can be said to be behavior based on impulsive judgment or behavior based on the current maximum reward, etc.

[0005] The information processing device of the first invention includes an NN storage unit storing neural network information having two or more pieces of node information each having a node identifier for identifying a node and corresponding to an area identifier for identifying an area of ​​the node, and one or more pieces of edge information each having an edge identifier for identifying an edge and specifying a connection between the nodes; a starting point storage unit storing one or more pieces of starting point information each having one or more node identifiers for identifying a node that will fire without passing through other nodes, the starting point condition being a condition related to a feature set that is one or more pieces of feature information of received information, and being a condition for determining a node that will fire without passing through other nodes; a firing condition storage unit storing firing conditions related to the feature set, the firing conditions being firing conditions related to one or more pieces of node information out of the two or more pieces of node information; an action management unit storing action information corresponding to two or more action identifiers for identifying actions; a cortex unit storing one or more pieces of long-term memory information each having a firing pattern each having one or more firing node identifiers that identify a node to fire, a past action identifier that identifies an action performed in the past, and an expected value of the action; a hippocampus unit storing one or more pieces of short-term memory information having one or more short-term action identifiers and action expectations corresponding to the short-term action identifiers; a reception unit receiving information; a feature acquisition unit acquiring one or more feature sets from the information received by the reception unit; a transmission unit performing a transmission process for determining one or more node identifiers from the start point storage unit that pair with start point conditions that match the feature sets acquired by the feature acquisition unit, and determining firing node identifiers of nodes that are connected by edges to fired nodes identified by the one or more node identifiers, are passed all or part of the feature sets, and will fire based on the firing conditions corresponding to the nodes; a long-term firing pattern having one or more firing node identifiers determined by the transmission unit and one or more firing node identifiers that pair with area identifiers indicating cortex areas; a firing pattern that satisfies a similarity condition for the long-term firing pattern and that is paired with the firing pattern stored in the cortex unit; and a short-term firing pattern having one or more firing node identifiers determined by the transmission unit and one or more firing node identifiers that pair with area identifiers indicating hippocampus areas.The information processing device includes a corpus call osmosis unit that acquires a score based on an expected value paired with the firing pattern stored in the hippocampus unit, which is a firing pattern that satisfies a similarity condition for the short-term firing pattern, acquires an action identifier that is a past action identifier or a short-term action identifier for a score that satisfies the action condition from the acquired two or more scores, acquires action information paired with the action identifier from the action management unit, and performs an action according to the action information.

[0006] This configuration makes it possible to simulate the movements in the brain when performing actions that take into account long-term memory and short-term memory.

[0007] In addition, the information processing device of the second invention is an information processing device that, compared to the first invention, the long-term memory information has time information regarding when the information was accumulated, and further includes a storage unit in which a first time condition, which is a condition regarding when the long-term memory information is applied and is a condition regarding the time information, is stored, and the corpus call osmosis unit acquires long-term firing patterns having one or more firing node identifiers determined by the transmission unit from among long-term firing patterns paired with time information that satisfies the first time condition, and which have one or more firing node identifiers paired with area identifiers indicating cortex regions.

[0008] This configuration makes it possible to more appropriately simulate the movements in the brain when performing actions that take into account long-term memory and short-term memory.

[0009] In addition, the information processing device of the third invention is an information processing device that, compared to the second invention, further includes a time condition change unit that obtains the number of ignition node identifiers determined by the transmission unit and changes the first time condition using the number.

[0010] This configuration makes it possible to more appropriately simulate the movements in the brain when performing actions that take into account long-term memory and short-term memory.

[0011] Furthermore, the information processing device of the fourth invention is an information processing device in which, compared to any one of the first to third inventions, the short-term memory information has time information regarding the time when the information was accumulated, a second time condition is stored which is a condition for applying the short-term memory information and is a condition regarding the time information, and the corpus call OSAM unit acquires short-term firing patterns which are one or more firing node identifiers determined by the transmission unit and which are paired with an area identifier indicating a hippocampus region from among the short-term firing patterns which are paired with the time information that satisfies the second time condition.

[0012] This configuration makes it possible to more appropriately simulate the movements in the brain when performing actions that take into account long-term memory and short-term memory.

[0013] In addition, the information processing device of the fifth invention is an information processing device that, compared to the fourth invention, further includes a time condition change unit that obtains the number of ignition node identifiers determined by the transmission unit and uses that number to change the second time condition.

[0014] This configuration makes it possible to more appropriately simulate the movements in the brain when performing actions that take into account long-term memory and short-term memory.

[0015] In addition, the information processing device of the sixth invention is an information processing device according to any one of the first to fifth inventions, wherein the reception unit receives new information after the corpus call OSAM unit has performed an action corresponding to the action information, and further includes a positive-amygdala unit that uses the firing pattern when the reception unit receives the new information to acquire the number of fired nodes among one or more nodes corresponding to an area identifier indicating a positive-amygdala region and acquires a positivity level using the number of nodes, and a negative-amygdala unit that uses the firing pattern when the reception unit receives the new information to acquire the number of fired nodes among one or more nodes corresponding to an area identifier indicating a negative-amygdala region and acquires a negativity level using the number of nodes, and the corpus call OSAM unit acquires a reward related to the difference between the positivity level acquired by the positive-amygdala unit and the negativity level acquired by the negative-amygdala unit, and uses the reward to update an expected value corresponding to the action identifier that identifies the action performed by the corpus call OSAM unit.

[0016] With this configuration, the expected value when performing an operation can be updated.

[0017] Furthermore, the information processing device of the seventh invention is an information processing device in which, compared to any one of the first to sixth inventions, the corpus call OSAM unit accumulates short-term memory information in the hippocampus unit, the short-term memory information having an action identifier that identifies an action performed by the corpus call OSAM unit and an firing pattern when the action is performed.

[0018] Such a configuration allows for the accumulation of short-term memory information including action identifiers of actions that have been performed.

[0019] Furthermore, the information processing device of the eighth invention is an information processing device in which, compared to any one of the first to seventh inventions, the Corpus Call OSAM unit updates the frequency of the action in association with an action identifier that identifies the action performed by the Corpus Call OSAM unit, and if the frequency satisfies the long-term memory condition, the Corpus Call OSAM unit accumulates long-term memory information in the cortex unit that has the action identifier that identifies the action performed by the Corpus Call OSAM unit and the firing pattern when the action is performed, and deletes short-term memory information that includes the action identifier from the hippocampus unit.

[0020] Such an arrangement allows for the accumulation of long-term memory information including action identifiers for actions that have been performed.

[0021] In addition, the information processing device of the ninth invention is an information processing device according to the sixth invention, further comprising: a survival rule storage unit in which one or more survival rules corresponding to a feature set are stored, the survival rule storage unit being a positive rule corresponding to positive information indicating positive or a negative rule corresponding to negative information indicating negative, and a rule application unit in which, when the judgment information is positive information, the transmission unit increases the weight of the edge connecting a node that is a node ahead of the fired node and corresponds to the area identifier of the positive area, and when the judgment information is negative information, the transmission unit increases the weight of the edge connecting a node that is a node ahead of the fired node and corresponds to the area identifier of the negative area.

[0022] This configuration allows for the construction of an appropriate neural network using managed survival rules.

[0023] The information processing device according to the present invention can simulate the activity in the brain when performing an action that takes into account long-term memory and short-term memory.

[0024] a block diagram of an information processing device 1 according to a first embodiment; a flowchart illustrating an example of an operation of the information processing device 1; a flowchart illustrating an example of a simultaneous firing transmission process; a flowchart illustrating an example of a start node process; a flowchart illustrating an example of a simultaneous firing time process; a flowchart illustrating an example of a simultaneous transmission process; a flowchart illustrating an example of a simultaneous firing determination process; a flowchart illustrating an example of a survival rule application process; a flowchart illustrating an example of a weight change process; a flowchart illustrating an example of a firing pattern process; a flowchart illustrating an example of a motion determination process; a flowchart illustrating an example of a long-term memory reference process; a flowchart illustrating an example of a short-term memory reference process; a flowchart illustrating an example of a post-processing process; a flowchart illustrating an example of a degree of similarity acquisition process; a flowchart illustrating an example of a simultaneous condition change process; a block diagram of a computer system;

[0025] Hereinafter, embodiments of an information processing device and the like will be described with reference to the drawings. Note that components with the same reference numerals in the embodiments perform similar operations, and therefore repeated description may be omitted.

[0026] First Embodiment In this embodiment, an information processing device will be described that includes a cortex section corresponding to long-term memory and a hippocampus section corresponding to short-term memory, and that can perform operations based on long-term memory and short-term memory.

[0027] In addition, in this embodiment, an information processing device capable of changing the time condition (Δt) in long-term memory and short-term memory will be described.

[0028] In this embodiment, an information processing device that can appropriately update an expected value corresponding to an action identifier will be described.

[0029] In addition, in this embodiment, an information processing device that can store action identifiers and the like of executed actions as short-term memory will be described.

[0030] In this embodiment, an information processing device will be described that stores action identifiers and the like as long-term memory when a condition is met.

[0031] Furthermore, in this embodiment, an information processing device capable of constructing an appropriate neural network using managed survival rules will be described.

[0032] In this specification, information X being associated with information Y means that information Y can be obtained from information X, or information X can be obtained from information Y, and the method of association is not important. Information X and information Y may be linked, may exist in the same buffer, information X may be included in information Y, or information Y may be included in information X, etc.

[0033] Furthermore, in this specification, selecting or determining information Z means obtaining information Z, obtaining a pointer to information Z, obtaining the ID of information Z, setting a flag on information Z, etc., and it is sufficient if information Z can be accessed.

[0034] 1 is a block diagram of an information processing device 1 according to the present embodiment. The information processing device 1 includes a storage unit 11, a reception unit 12, a processing unit 13, and an output unit 14. The storage unit 11 includes an NN storage unit 111, a starting point storage unit 112, a firing condition storage unit 113, an operation management unit 114, a cortex unit 115, a hippocampus unit 116, and a survival rule storage unit 117.

[0035] The receiving unit 12 includes an information receiving unit 121. The processing unit 13 includes a feature acquiring unit 131, a transmitting unit 132, a corpus call summing unit 133, a positive / negative determination unit 134, a negative / negative determination unit 135, a time condition changing unit 136, a judgment unit 137, and a rule applying unit 138. The corpus call summing unit 133 includes a pattern acquiring means 1331, an action determining means 1332, and an action means 1333.

[0036] The information processing device 1 is, for example, a terminal, but may also be a server. When the information processing device 1 is a terminal, the information processing device 1 is, for example, a personal computer, a tablet terminal, or a smartphone, but may also be a robot, etc. When the information processing device 1 is a robot, the robot may be a human-shaped robot. When the information processing device 1 is a server, the information processing device 1 is, for example, a cloud server or an ASP server, but this does not matter. When the information processing device 1 is a server, for example, there is a terminal that provides information to the information processing device 1.

[0037] Various types of information are stored in the storage unit 11 constituting the information processing device 1. The various types of information include, for example, neural network information (hereinafter referred to as "NN information"), starting point information, firing conditions, operation information, firing information, and various conditions. The various conditions include, for example, first-time conditions, second-time conditions, similarity conditions, operation conditions, and long-term memory conditions. It goes without saying that the various conditions may be embedded in the program.

[0038] Firing information is information about the result of firing. The firing information has a node identifier that identifies the node that fired. The firing information may usually have timer information that indicates the time of firing. The timer information may be information that indicates relative time or time information that indicates absolute time. The timer information may be considered to be a node attribute value. The firing information may have a node identifier of the node from which the firing is transmitted. The firing information may have an edge identifier of the edge when the firing is transmitted. Note that the firing information may be automatically deleted by the processing unit 13 after a certain time has passed since it was accumulated.

[0039] The temporary condition is a condition regarding the time when the long-term memory information is applied, and is a condition regarding the time information. An example of the temporary condition is "Δt<=10 years." "Δt" is the time that has elapsed since the long-term memory information was accumulated. The time information is information regarding the time when the information was accumulated. The time information usually includes the date on which the information was accumulated. The time information is the date on which the information was accumulated, or the date and time when the information was accumulated.

[0040] The second time condition is a condition to which the short-term memory information is applied, and is a condition related to time information. For example, the second time condition is "Δt<=24 hours." Here, "Δt" is the elapsed time since the short-term memory information was accumulated.

[0041] The NN storage unit 111 stores neural network information (hereinafter referred to as "NN information" where appropriate). The NN information is information that simulates a neural network (hereinafter referred to as "NN" where appropriate) in the brain. The NN information is information that identifies a NN having two or more nodes and one or more edges connecting the nodes. The NN information has two or more node information and one or more edge information. It is preferable that each node of the NN information is connected to one or more other nodes by an edge. In other words, it is preferable that the NN here does not have an input layer or an output layer.

[0042] Node information is information about the nodes that make up the NN. The node information has a node identifier. The node information has, for example, one or more node attribute values. Each of the two or more pieces of node information that make up the NN information corresponds to an area identifier. The area identifier is information that identifies an area in the brain. The area identifier is, for example, an area name or an area ID. The area identifier is, for example, "cortex," "hippocampus," "positivity," or "negativity." It is preferable that the area identifier is information that identifies one of four or more areas including "cortex," "hippocampus," "positivity," or "negativity." The "cortex" area is an area of ​​long-term memory. The "hippocampus" area is an area of ​​short-term memory. One or more pieces of node information among the two or more pieces of node information may be associated with two area identifiers: "positivity" and "negativity." The "positive" area is the area of ​​the node that fires in the positive case. The "negative" area is the area of ​​the node that fires in the negative case. Note that the NN information may include node information that does not correspond to an area identifier.

[0043] A node identifier is information that identifies a node that constitutes a network. A node identifier is, for example, a node ID or node name. A node may also be called a soma. A node identifier may also be called a soma identifier.

[0044] The node attribute values ​​include, for example, node position information, firing conditions, firing probability information, and number of times information.

[0045] The node position information is information about the position of a node, and is, for example, three-dimensional coordinate values ​​(x, y, z), two-dimensional coordinate values ​​(x, y), or four-dimensional quaternion (x, y, x, w).

[0046] The firing condition is a condition under which a node fires. The firing condition may be considered to be the firing condition stored in the firing condition storage unit 113. The firing condition will be described in detail later.

[0047] The firing probability information is information relating to the probability of firing. The firing probability information may be the firing probability itself, or a value obtained by converting the firing probability using a function or the like. It is preferable that the firing probability information is referenced, and the node may or may not fire at the probability specified by the firing probability information, even if the feature information corresponding to the node is the same.

[0048] The number of times information is information based on the number of times a node has fired. The number of times information is, for example, the number of times fired or the firing rate. The number of times fired is, for example, the number of times fired in a specific period, or the cumulative number of times fired. The firing rate is, for example, the rate of firing in a specific period, or the rate of firing in the entire period. The firing rate of a node is the "number of times fired / number of times a signal was transmitted to that node."

[0049] Edge information is information about the edges that make up a neural network. Edge information is information that identifies edges that connect nodes. Edge information typically includes an edge identifier. Edge identifiers are information that identify edges. The edge identifier is, for example, the edge ID or the edge name. Edge information is, for example, the node identifier of each of the two nodes that the edge connects. Edge information is, for example, the node identifier of one connecting node. When edge information includes only one node identifier, the edge is in the growth process before connecting the two nodes. Edge information includes one or more edge attribute values. The edge attribute values ​​are, for example, weight and edge position information. Weight can be considered information based on the thickness of the edge. The larger the weight, the more easily information is transmitted via the edge. The larger the edge weight, typically, the more likely the node beyond the edge is to fire. Edge position information is edge position information. Edge position information is information that identifies the position of the endpoint of the edge. When edge information includes two node identifiers, for example, edge position information is not included. When edge information has two node identifiers, this means that the edge specified by the edge information connects two nodes.

[0050] The edge information includes, for example, dendrites information and AXON information. In this case, the edge includes dendrites and AXON. Note that the edge may be considered to be a single line or two or more branched lines. The edge may also be called a synapse.

[0051] Dendrites information is information about DENDRITES. DENDRITES are also called dendrites and are parts of nerve cells. They are multiple projections that branch out from the cell body like the branches of a tree in order for nerve cells to receive external stimuli and information sent from the axons (AXONs) of other nerve cells. DENDRITES are elements that make up edges in this case. DENDRITES information includes a DENDRITES identifier and DENDRITES position information.

[0052] The DENDRITES identifier is information for identifying DENDRITES, such as the ID or name of DENDRITES.

[0053] DENDRITES position information is information about the position of DENDRITES. DENDRITES position information is information that specifies the position of DENDRITES, and is, for example, one or more three-dimensional coordinate values ​​(x, y, z) or one or more two-dimensional coordinate values ​​(x, y). When the DENDRITES position information has two or more coordinate values, DENDRITES is a line connecting the points of the two or more coordinate values.

[0054] It is also preferable that the dendrites information includes information on the amount of energy held by the dendrites, which indicates the amount of energy held by the dendrites.It is also preferable that the dendrites information includes information on the amount of energy required to transmit information using the dendrites.

[0055] The axon information is information about an axon. An axon, also known as an axon, is a protruding structure extending from a cell body and responsible for signal output in a nerve cell. Here, an axon is an element that constitutes an edge. The axon information includes an axon identifier and axon position information.

[0056] The AXON identifier is information for identifying the AXON, such as the ID or name of the AXON.

[0057] AXON position information is information about the position of the AXON. The AXON position information is information that specifies the position of the AXON, and is, for example, one or more three-dimensional coordinate values ​​(x, y, z) or one or more two-dimensional coordinate values ​​(x, y). When the AXON position information has two or more coordinate values, the AXON is a line connecting the points of the two or more coordinate values.

[0058] Preferably, the AXON information includes information on the amount of energy stored in the AXON, and information on the amount of energy required to transmit information using the AXON.

[0059] Note that dendrites and AXON may be branched. When dendrites and AXON are branched, the position information of each may be expressed by three or more coordinate values. However, the method of expressing dendrite position information and AXON position information does not matter.

[0060] The start point storage unit 112 stores one or more pieces of start point information. The start point information is information for determining the node that will fire first when information is received. The start point information corresponds to a start point condition. The start point information usually has a node identifier that identifies the node that can fire first. A node that can fire first is a node that does not receive information from other fired nodes. A node that can fire first is, for example, a node in the input layer. The start point information has, for example, one or more node identifiers and a start point condition. A start point condition is a condition for determining a node that will fire without going through other nodes. A start point condition is a condition related to one or more pieces of feature information obtained from the information received by the information receiving unit 121. Note that one or more pieces of feature information are referred to as a "feature set" as appropriate.

[0061] The firing condition storage unit 113 stores one or more firing conditions. The firing condition corresponds to the node information. The firing condition may be included in the node information. The firing condition is a condition for a node to fire. The firing condition is a condition related to a feature set.

[0062] The firing condition storage unit 113 stores firing conditions corresponding to one or more pieces of node information among the two or more pieces of node information. Note that node information not corresponding to a firing condition may or may not exist. Node information not corresponding to a firing condition is information about a node that will always fire when a signal is transmitted.

[0063] The firing condition is, for example, a condition related to one or more feature information of the sound information. In such a case, the firing condition is called a sound feature condition. The sound feature condition is, for example, a condition related to a specific frequency of the sound information. In such a case, the firing condition is called a frequency condition. The frequency condition is, for example, information specifying a frequency range (e.g., "first threshold <= frequency <= second threshold"), information specifying an upper frequency limit (e.g., "frequency <= second threshold"), or information specifying a lower frequency limit (e.g., "first threshold <= frequency"). Examples of frequency conditions are "<information identifier> frequency = Fa <condition> information amount > = 100", "<information identifier> frequency = Fa <condition> information amount > = 100", or "<information identifier> frequency = Fa <condition> information amount > = 150 & <information identifier> frequency = Fb <condition> information amount > = 180". Note that the frequencies Fa and Fb are values ​​indicating specific frequencies. Also, "<information identifier> frequency=Fa <condition> information amount>=100" means that the information amount (for example, volume) of the frequency "Fa" is 100 or more.

[0064] The firing condition is, for example, a condition related to one or more feature information of the image information. Such a firing condition is called an image feature condition. An example of the image feature condition is "<information identifier>R <condition> information amount >= 150". An example of the initial firing condition is "information amount of 'R' >= 150", "(information amount of 'R' >= 150) & (information amount of 'G' >= 80)", or "(information amount of 'R' >= 150) & (information amount of 'G' >= 80) & (information amount of 'B' >= 120)".

[0065] The firing condition is, for example, a condition related to the frequency at which a signal is transmitted to a node. Such a firing condition is called a frequency condition. The frequency condition is a condition related to frequency information. Frequency information is information related to the frequency at which information is transmitted to a node. Frequency information is, for example, the time resolution at which a signal is transmitted or the number of times a signal is transmitted within a predetermined time. Note that the information here is typically one or more characteristic pieces of information about the received information. The information here may also be referred to as a signal. Frequency conditions are, for example, a condition for a specific time resolution, a condition for a specific range of time resolution, a condition for specific rhythm information, a condition for a range of similarity with a reference specific rhythm information, etc. The time resolution is the number of times a signal fires per unit time. The rhythm information is a vector whose elements are the values ​​of two or more intervals (times) during which a signal is transmitted. Note that the transmission of a signal to a node means, for example, that another node connected to the node by an edge fires. When a signal is transmitted to a node, the node in question is, for example, a node connected to the edge determined to transmit the signal.

[0066] The firing condition may be, for example, a combination of two or more conditions selected from a sound feature condition, an image feature condition, and a frequency condition.

[0067] The operation management unit 114 stores two or more pieces of operation information. An operation identifier corresponds to each of the two or more pieces of operation information. Operation information is information that identifies an operation. The operation information is, for example, a program, a module, an operation name, etc. The operation information may have a positive or negative degree. Such a positive or negative degree may not be fixed, but may change dynamically. An operation identifier is information that identifies an operation. The operation identifier is, for example, an operation ID, an operation name (for example, a function name, a method name, a module name), etc.

[0068] The cortex unit 115 is a component responsible for long-term memory. The cortex unit 115 stores one or more pieces of long-term memory information. Long-term memory information is information that is stored for a long period of time. The long-term memory information includes a firing pattern, a past action identifier, and an expected value. A firing pattern is information that identifies one or more fired nodes. A firing pattern includes one or more fired node identifiers. A fired node identifier is information that identifies a fired node. A fired node is a node that has fired. A past action identifier is an action identifier that identifies a past action. A past action identifier is usually an action identifier that identifies an action performed a predetermined period of time or more ago. An action identifier is information that identifies a performed action. Examples of past action identifiers are "laughing" and "crying." The expected value can be said to be information that identifies a reward that will be obtained when an action identified by a paired action identifier is performed. The reward can also be called a benefit, etc. The cortex unit 115 can also be called "Cortex 115."

[0069] The hippocampus section 116 is a component that controls short-term memory. The hippocampus section 116 stores one or more pieces of short-term memory information. Short-term memory information is information that is stored only for a short period of time. The short-term memory information includes a firing pattern, a short-term action identifier, and an expected value. The short-term action identifier is an action identifier that identifies an action that has been performed recently. "Recent" usually refers to a predetermined period of time. The predetermined period is a predetermined period from the current point in time. Note that the predetermined period may change. An example of a short-term action identifier is "raise hand." Note that the hippocampus section 116 may also be called "hippocampus116" or "HIP116."

[0070] When a predetermined period of time has elapsed since the action identified by the short-term action identifier of the hippocampus unit 116 was performed, for example, the short-term memory information having the short-term action identifier is deleted from the hippocampus unit 116. The deleted short-term memory information may be written to the cortex unit 115 as long-term memory information.

[0071] The survival rule storage unit 117 stores one or more survival rules. Each of the one or more survival rules is associated with judgment information. The judgment information is information that specifies the type of survival rule. The judgment information is positive information or negative information. Positive information is information that indicates a positive result, such as "1". Negative information is information that indicates a negative result, such as "0". Each of the one or more survival rules may be associated with both positive information and negative information. Each of the one or more survival rules is associated with one or more feature sets. A feature set is one or more pieces of feature information.

[0072] A survival rule is, for example, a rule used by a person to survive. Each of the one or more survival rules is a positive rule or a negative rule. A positive rule is a rule that defines something as positive. A positive rule is a rule that corresponds to positive information. A negative rule is a rule that defines something as negative. A negative rule is a rule that corresponds to negative information. Here, rules are usually information. A positive rule is, for example, a feature set of images of delicious food or a feature set of laughter. A negative rule is, for example, a feature set of images of cockroaches.

[0073] The receiving unit 12 receives various types of information and instructions. The various types of information and instructions include, for example, images, voice, sounds, character strings, and operation instructions. An operation instruction is an instruction to start an operation.

[0074] The receiving unit 12 receives the information after the corpus call processing unit 133 performs an operation according to the operation information. Such information is new information.

[0075] The information receiving unit 121 receives information, such as an image, a voice, a sound, or a character string.

[0076] Here, reception is a concept that includes, for example, reception of information input from an input device such as a keyboard, mouse, or touch panel, acquisition by a microphone, acquisition by a camera, reception of information transmitted via a wired or wireless communication line, and reception of information read from a recording medium such as an optical disk, magnetic disk, or semiconductor memory.

[0077] The means for inputting information may be any means, such as a touch panel, keyboard, mouse, menu screen, microphone, or camera.

[0078] The processing unit 13 performs various processes, such as processes performed by a feature acquisition unit 131, a transmission unit 132, a corpus call analysis unit 133, a positive analysis unit 134, a negative analysis unit 135, a time condition change unit 136, a judgment unit 137, or a rule application unit 138.

[0079] The feature acquisition unit 131 acquires a feature set, which is one or more pieces of feature information, from the information received by the reception unit 12. The feature information may also be referred to as a feature amount.

[0080] When the received information is sound information, the acquired feature information is called sound feature information. It is preferable that the one or more pieces of sound feature information include frequency. The one or more pieces of sound feature information may include volume. The one or more pieces of sound feature information may be, for example, MFCC or Mel spectrum, but this is not important. When the received information is image information, the acquired feature information is called image feature information. The one or more pieces of image feature information are typically R, G, and B values. The one or more pieces of image feature information may include brightness, lightness, and saturation. When the received information is a character string, the acquired feature information is, for example, one or more independent words.

[0081] The transmission unit 132 determines one or more node identifiers corresponding to the start point condition that the feature set acquired by the feature acquisition unit 131 matches. The start point condition that the feature set matches is, for example, a start point condition that one or more feature information items included in the feature set match. This determination is to determine the node that will fire first when the information is accepted. Determining a node identifier, for example, is to obtain a node identifier. Determining a node identifier is to cause a node identified by the node identifier to fire. Note that when a node fires, conceptually, for example, one or more pieces of feature information are passed to other nodes connected to the node by edges. These one or more pieces of feature information are information that the node at least temporarily holds. Note that the transmission unit 132 may pass only a portion of the one or more pieces of feature information passed to a node to other nodes connected to the node by edges.

[0082] The transmission unit 132 is a node that is connected to the fired node by an edge, receives one or more pieces of feature information from the fired node, and determines the node identifier of the node that will fire based on the firing condition corresponding to the node. This processing is a firing transmission processing.

[0083] It is preferable that the transmission unit 132 recursively determines the node to be fired and acquires the node identifier of the node.

[0084] The transmitter 132 is a node connected to the fired node by an edge, and when determining the node identifier of a node that will fire based on the firing condition corresponding to the node, it is preferable to perform processing such that the greater the weight of the edge, the more likely the node is to be determined as a node that will fire. Note that the processing for making it easier to determine a node is, for example, using the weight of the edge to determine that the greater the weight, the more likely the node is to fire. The processing for making it easier to determine a node is, for example, processing for making it easier to determine a node is, for example, processing for making the firing probability information included in the node information of a node connected to the edge greater, the greater the weight of the edge.

[0085] After determining the node to be fired, the transmission unit 132, for example, obtains timer information that specifies the time of firing, composes firing information having the node identifier of the fired node and the timer information, and stores the firing information in the storage unit 11.

[0086] The pattern acquisition unit 1331 included in the corpus call analysis unit 133 acquires a firing pattern. The firing pattern is information about one or more firing nodes. The firing pattern includes, for example, one or more node identifiers determined by the transmission unit 132.

[0087] It should be noted that there is no restriction on the timing at which the pattern acquisition unit 1331 acquires the firing pattern. For example, the pattern acquisition unit 1331 acquires the firing pattern when a node fires, periodically, constantly, or when there is a node that has fired within a predetermined time from the current time.

[0088] The pattern acquisition means 1331 acquires, for example, a firing pattern that can be acquired from the firing information in the storage unit 11. The pattern acquisition means 1331 acquires, for example, a firing pattern that includes a node identifier held by one or more pieces of firing information in the storage unit 11. Such firing information is, for example, firing information that has been stored in the storage unit 11 and has not been deleted. Such firing information is, for example, firing information that has been stored in the storage unit 11 for a predetermined period of time. If the NN has an output layer, such firing information is, for example, information on fired nodes among the nodes in the output layer.

[0089] The pattern acquisition means 1331 acquires, for example, from among the long-term firing patterns paired with time information that satisfies the first time condition, a long-term firing pattern that has one or more firing node identifiers determined by the transmission unit 132 and that pair with an area identifier indicating a cortex region.

[0090] The pattern acquisition means 1331 acquires, for example, from among the short-term firing patterns paired with time information that satisfies the second time condition, one or more firing node identifiers determined by the transmission unit 132, and one or more firing node identifiers paired with an area identifier indicating a hippocampus region.

[0091] The action determination means 1332 is a firing pattern that satisfies the similarity condition for the long-term firing pattern acquired by the pattern acquisition means 1331 and acquires a score based on an expected value paired with the firing pattern stored in the cortex unit 115. The action determination means 1332 also is a firing pattern that satisfies the similarity condition for the short-term firing pattern acquired by the pattern acquisition means 1331 and acquires a score based on an expected value paired with the firing pattern stored in the hippocampus unit 116. Next, the action determination means 1332 acquires an action identifier that is a past action identifier or a short-term action identifier for a score that satisfies the action condition from the two or more acquired scores. The action determination means 1332 may acquire two or more action identifiers.

[0092] The similarity between two firing patterns is, for example, the similarity between the vectors that are the two firing patterns. The similarity between two firing patterns is, for example, the number of node identifiers that are common to the two firing patterns. There is no restriction on the method for calculating the similarity between two firing patterns. The similarity condition is usually "the similarity is equal to or greater than a threshold value."

[0093] The score based on the expected value may be the expected value itself, or may be a value obtained using the expected value and the similarity between two firing patterns. The score increases as the expected value increases, and as the similarity between two firing patterns increases.

[0094] For example, when the information processing device 1 imitates a person who makes calm decisions or a person who makes decisions based on past cases, the action determination means 1332 obtains an expected value that is a firing pattern that satisfies the similarity condition with respect to the long-term firing pattern acquired by the pattern acquisition means 1331 and that pairs with the firing pattern stored in the cortex unit 115, and uses the expected value to obtain a higher score. On the other hand, when the information processing device 1 imitates a person who makes impulsive decisions or a person who makes decisions based on the current maximum reward, the action determination means 1332 obtains an expected value that is a firing pattern that satisfies the similarity condition with respect to the long-term firing pattern acquired by the pattern acquisition means 1331 and that pairs with the firing pattern stored in the cortex unit 115, and uses the expected value to obtain a lower score.

[0095] For example, when the information processing device 1 imitates a person who makes calm decisions or a person who makes decisions based on past cases, the action determination means 1332 acquires an expected value that is a firing pattern that satisfies the similarity condition with respect to the short-term firing pattern acquired by the pattern acquisition means 1331 and that pairs with the firing pattern stored in the hippocampus unit 116, and uses the expected value to acquire a lower score. On the other hand, when the information processing device 1 imitates a person who makes impulsive decisions or a person who makes decisions based on the current maximum reward, the action determination means 1332 acquires an expected value that is a firing pattern that satisfies the similarity condition with respect to the short-term firing pattern acquired by the pattern acquisition means 1331 and that pairs with the firing pattern stored in the hippocampus unit 116, and uses the expected value to acquire a higher score.

[0096] The operating conditions are, for example, "the score is the maximum," "the score is equal to or greater than a threshold," "the score is in the top N (N is a natural number equal to or greater than 1)," and "the score is Nth (N is a natural number equal to or greater than 1)." The operating conditions may be fixed or may change dynamically.

[0097] The action determination means 1332 may, for example, randomly determine one action identifier from action identifiers that are past action identifiers or short-term action identifiers for each of the two or more acquired scores. In such a case, the randomly determined one action identifier is the action identifier that satisfies the action condition.

[0098] The action determination means 1332 may, for example, sort the action identifiers, which are past action identifiers or short-term action identifiers for each of the two or more acquired scores, in order of score, and determine one action identifier for the acquired ranking based on a specific probability for each ranking. In such a case, the determined one action identifier is the action identifier that satisfies the action condition.

[0099] The operation means 1333 acquires operation information paired with one or more operation identifiers acquired by the operation determination means 1332 from the cortex section 115 or the hypocampus section 116, and performs an operation according to the one or more operation information.

[0100] The operation according to the operation information is the operation specified by the operation information. For example, performing the operation according to the operation information means executing a program or module included in the operation information, or performing an operation identified by an operation identifier associated with the operation information.

[0101] The corpus call summing unit 133 obtains a reward related to the difference between the positivity obtained by the positive assessment unit 134 and the negativity obtained by the negative assessment unit 135. Then, the corpus call summing unit 133 preferably uses the reward to update an expectation value corresponding to an action identifier that identifies the action performed by the corpus call summing unit 133.

[0102] When the corpus call OSAM unit 133 updates the expected value using the reward, it may be by adding a new expected value or by updating the expected value associated with the action identifier.

[0103] When a new expected value is added, the corpus call OSAM unit 133 uses the acquired reward to acquire a new expected value and associates the new expected value with the action identifier. In this case, it is preferable that the corpus call OSAM unit 133 associates the firing pattern acquired by the pattern acquisition means 1331 with the action identifier in addition to the acquired new expected value. The new expected value to be added may be the reward, or may be a value calculated by an increasing function using the reward as a parameter.

[0104] When updating an expected value associated with an action identifier, the corpus call OSAM unit 133, for example, adds a reward to the expected value associated with the action identifier. The corpus call OSAM unit 133, for example, calculates a new expected value using an increasing function with the expected value associated with the action identifier and the reward as parameters, and associates the new expected value with the action identifier.

[0105] When updating an expected value associated with a behavior identifier, the corpus call OSAM unit 133 may also update, for example, the firing pattern associated with the behavior identifier. For example, the corpus call OSAM unit 133 overwrites the original firing pattern with the firing pattern acquired by the pattern acquisition unit 1331.

[0106] The reward may be the difference between the positivity and negativity, or a value obtained by an increasing function that uses the difference between the positivity and negativity as a parameter, or a value obtained by a function that is an increasing function that uses the positivity as a parameter and a decreasing function that uses the negativity as a parameter. The greater the positivity, the greater the value of the reward, and the greater the negativity, the smaller the value.

[0107] The corpus call OSAM unit 133 stores, in the hippocampus unit 116, short-term memory information having, for example, an action identifier that identifies an action performed by the corpus call OSAM unit 133 and an ignition pattern when the action is performed.

[0108] The corpus call summing unit 133 updates the frequency of the action, for example, in association with an action identifier that identifies the action performed by the corpus call summing unit 133 .

[0109] If the frequency satisfies the long-term memory condition, the corpus call OSAM unit 133 stores long-term memory information having an action identifier that identifies the action performed by the corpus call OSAM unit 133 and an ignition pattern when the action is performed in the cortex unit 115. In addition, the corpus call OSAM unit 133 deletes the short-term memory information including the action identifier from the hippocampus unit 116.

[0110] The positivity calculator 134 uses the firing pattern when the receiving unit 12 receives information to obtain the number of fired nodes among one or more nodes corresponding to the area identifier indicating the positivity, and obtains the positivity level using the number of nodes. The positivity calculator 134 obtains a higher positivity level the greater the number of nodes. The positivity calculator 134 calculates the positivity level, for example, using an increasing function with the number of nodes as a parameter. The positivity calculator 134 obtains the positivity level corresponding to the number of nodes, for example, from a correspondence table. The correspondence table has two or more pieces of correspondence information each having a range of the number of nodes and a positivity level.

[0111] Preferably, the negativity unit 135 acquires the number of fired nodes among one or more nodes corresponding to the area identifier indicating the negativity using the firing pattern when the receiving unit 12 receives information, and acquires the negativity degree using the number of fired nodes. The larger the number of nodes, the larger the negativity degree acquired by the negativity unit 135. For example, the negativity unit 135 calculates the positivity degree using an increasing function with the number of nodes as a parameter. For example, the negativity unit 135 acquires the negativity degree corresponding to the number of nodes from a correspondence table. The correspondence table has two or more pieces of correspondence information each having a range of the number of nodes and a negativity degree.

[0112] The time condition changing unit 136 obtains the number of ignition nodes, which is the number of ignition node identifiers determined by the transmission unit 132, and changes the first time condition using the number of ignition nodes. The time condition changing unit 136 changes the first time condition to a longer time as the number of ignition nodes increases.

[0113] The time condition changing unit 136 may determine whether the number of firing nodes satisfies a first change condition, and change the first time condition only if the first change condition is satisfied. The first change condition may be, for example, "the number of firing nodes is equal to or greater than a threshold" or "the difference between the number of firing nodes last time and the number of firing nodes this time is equal to or greater than a threshold."

[0114] The time condition change unit 136 may obtain the number of firing nodes in the cortex area, which is the number of node identifiers that pair with the area identifier of the cortex area, from the firing node identifiers determined by the transmission unit 132, and use the number of firing nodes to change the first time condition.

[0115] The time condition changing unit 136 may determine whether the number of ignition nodes in the cortex region satisfies a first change condition, and change the first time condition only if the first change condition is satisfied.

[0116] Note that the time condition changing unit 136 normally changes the first time condition to a longer time as the number of ignition nodes in the cortex region increases.

[0117] The time condition changing unit 136 may obtain the number of ignition nodes, which is the number of ignition node identifiers determined by the transmission unit 132, and change the second time condition using the number of ignition nodes. The time condition changing unit 136 changes the second time condition to a longer time as the number of ignition nodes increases.

[0118] The time condition changing unit 136 may determine whether the number of firing nodes satisfies a second change condition, and change the second time condition only if the second change condition is satisfied. The second change condition may be, for example, "the number of firing nodes is equal to or greater than a threshold" or "the difference between the number of firing nodes last time and the number of firing nodes this time is equal to or greater than a threshold."

[0119] The time condition change unit 136 may obtain the number of firing nodes in the hippocampus region, which is the number of node identifiers that pair with the area identifier of the hippocampus region, from the firing node identifiers determined by the transmission unit 132, and change the second time condition using the number of firing nodes.

[0120] The time condition changing unit 136 may determine whether or not the number of ignition nodes in the hippocampus region satisfies the second change condition, and change the second time condition only if the second change condition is satisfied.

[0121] Note that the time condition changing unit 136 normally changes the second time condition to a longer time as the number of ignition nodes in the hippocampus region increases.

[0122] In addition, it is preferable that the acquired number of fired nodes is stored in the storage unit 11 in association with time information.

[0123] Furthermore, the timing or trigger for the time condition change unit 136 to change the first change condition or the second change condition is not important. The time condition change unit 136 performs processing to change the change condition periodically or constantly, for example, when a firing and transmission process in response to reception of a piece of information is completed, when any operation is executed, or the like.

[0124] The determination unit 137 determines a survival rule corresponding to the feature set acquired by the feature acquisition unit 131, and acquires determination information, which is positive information or negative information associated with the survival rule.

[0125] If the judgment information acquired by the judgment unit 137 is positive, the rule application unit 138 increases the weight of the edge connecting the node that is the node ahead of the node that was fired by the transmission unit 132 and that corresponds to the area identifier of the area of ​​the positive area. The amount by which the edge weight is increased does not matter.

[0126] If the determination information acquired by the determination unit 137 is negative information, the rule application unit 138 increases the weight of the edge connecting the node that is the node ahead of the node that was fired by the transmission unit 132 and that corresponds to the area identifier of the negative area. The amount by which the edge weight is increased does not matter.

[0127] The output unit 14 outputs various types of information. The various types of information are, for example, the results of the operations performed by the corpus call summing unit 133. Note that the corpus call summing unit 133 may output the results of the operations performed by the corpus call summing unit 133. In such a case, the output unit 14 is not necessary.

[0128] Output is a concept that includes displaying on a display, projecting using a projector, printing on a printer, outputting sound, transmitting to an external device, storing on a recording medium, and handing over processing results to other processing devices or other programs.

[0129] The storage unit 11, NN storage unit 111, starting point storage unit 112, firing condition storage unit 113, operation management unit 114, cortex unit 115, hypocampus unit 116, and survival rule storage unit 117 are preferably non-volatile recording media, but can also be realized using volatile recording media.

[0130] There is no restriction on the process by which information is stored in the storage unit 11 etc. For example, information may be stored in the storage unit 11 etc. via a recording medium, information transmitted via a communication line etc. may be stored in the storage unit 11 etc., or information input via an input device may be stored in the storage unit 11 etc.

[0131] The reception unit 12 and the information reception unit 121 are realized, for example, by wireless or wired communication means, means for receiving broadcasts, device drivers for input means such as a touch panel, microphone, camera, or keyboard, and control software for a menu screen.

[0132] The processing unit 13, feature acquisition unit 131, transmission unit 132, corpus call analysis unit 133, positive and negative analysis unit 134, negative and positive analysis unit 135, time condition change unit 136, judgment unit 137, rule application unit 138, pattern acquisition means 1331, action determination means 1332, and action means 1333 can typically be realized by a processor, memory, etc. The processing procedures of the processing unit 13, etc. are typically realized by software, and the software is recorded on a recording medium such as a ROM. However, they may also be realized by hardware (dedicated circuitry). The processor may be a CPU, MPU, GPU, etc., and the type does not matter.

[0133] The output unit 14 is realized by, for example, driver software for an output device such as a display or a speaker, or driver software for an output device and the output device.

[0134] Next, an example of the operation of the information processing device 1 will be described with reference to the flowchart of FIG.

[0135] (Step S201) The information receiving unit 121 determines whether or not information has been received. If information has been received, the process proceeds to step S202, and if information has not been received, the process proceeds to step S204.

[0136] (Step S202) The transmission unit 132 and the like perform an ignition transmission process. An example of the ignition transmission process will be described with reference to the flowchart in Fig. 3. The ignition transmission process is a process that imitates the transmission of a signal within a neural network in response to the reception of information.

[0137] (Step S203) The rule application unit 138 and the like perform a survival rule application process. Return to step S201. An example of the survival rule application process will be described using the flowchart in FIG. 8. The survival rule application process is a process of determining a survival rule corresponding to the received information and increasing the weight of the edge according to the determined survival rule.

[0138] (Step S204) The processing unit 13 determines whether it is time to perform firing pattern processing. If it is time to perform firing pattern processing, the process proceeds to step S205, and if it is not time to perform firing pattern processing, the process proceeds to step S206.

[0139] The timing for performing the firing pattern processing may be, for example, when a node fires, periodically, constantly, or when a node that has fired exists within a predetermined time from the current time. The timing for performing the firing pattern processing is not important. When a node fires, it is usually when the firing information is accumulated in the storage unit 11.

[0140] (Step S205) The corpus call OSAM unit 133 and the like perform firing pattern processing. Return to step S201. An example of the firing pattern processing will be described using the flowchart in FIG. 10. The firing pattern processing is processing such as detecting a firing pattern related to a fired node and executing an operation corresponding to the firing pattern.

[0141] (Step S206) The processing unit 13 determines whether it is time to change the time condition. If it is time to change the time condition, the processing proceeds to step S207, and if it is not time to change the time condition, the processing returns to step S201.

[0142] The time condition here is one or both of the first time condition and the second time condition. The timing for changing the time condition is, for example, when a firing and transmission process in response to reception of one piece of information is completed, when an operation is executed, periodically, constantly, etc. The timing for changing the time condition is not important.

[0143] (Step S207) The time condition change unit 136 performs the time condition change process. Return to step S201. An example of the time condition change process will be described with reference to the flowchart in FIG.

[0144] In the flowchart of FIG. 2, the process ends when the power is turned off or an interrupt occurs to end the process.

[0145] Next, an example of the firing transmission process in step S202 will be described with reference to the flowchart of FIG.

[0146] (Step S301) The transmission unit 132 assigns 1 to a counter i.

[0147] (Step S302) The transmission unit 132 determines whether the i-th information exists among the information received in S201. If the i-th information exists, the process proceeds to step S303; if not, the process proceeds to step S306.

[0148] The information received in S201 is, for example, one or two types of information from among audio information and image information.

[0149] (Step S303) The feature acquisition unit 131 acquires one or more pieces of feature information of the i-th information and temporarily stores them in a buffer (not shown). If the i-th information is audio information, it is preferable that the one or more pieces of feature information include frequency. If the i-th information is image information, it is preferable that the one or more pieces of feature information include the amount of information for each of "R", "G", and "B".

[0150] (Step S304) The transmission unit 132 increments the counter i by 1. The process returns to step S302.

[0151] (Step S305) The transmission unit 132 performs start node processing using one or more pieces of feature information stored in a buffer (not shown). An example of the start node processing will be described with reference to the flowchart in Fig. 4. The start node processing is processing for determining the node that will be fired first based on the received information.

[0152] (Step S306) The transmission unit 132 assigns 1 to the counter i.

[0153] (Step S307) The transmission unit 132 determines whether or not the i-th ignition node exists among the nodes determined to ignite in step S305. If the i-th ignition node exists, the process proceeds to step S308; if not, the process returns to the upper level process.

[0154] (Step S308) The transmission unit 132 performs next transmission processing. An example of the next transmission processing will be described using the flowchart in FIG. 6. The next transmission processing is processing that imitates the transmission of firing from one node to a next node connected to the first node by an edge. The next transmission processing is usually a processing that determines the node that will fire next, which is a node to which one or more pieces of feature information are passed from the firing node of interest.

[0155] (Step S309) The transmission unit 132 increments the counter i by 1. The process returns to step S307.

[0156] Next, an example of the start node process in step S305 will be described with reference to the flowchart of FIG.

[0157] (Step S401) The transmission unit 132 assigns 1 to a counter i.

[0158] (Step S402) The transmission unit 132 determines whether or not the i-th start point information exists in the start point storage unit 112. If the i-th start point information exists, the process proceeds to step S403; if not, the process returns to the upper level process.

[0159] (Step S403) The transmitter 132 acquires the start point condition included in the i-th start point information from the start point storage unit 112.

[0160] (Step S404) The transmission unit 132 determines whether or not one or more pieces of feature information stored in a buffer (not shown) in step S303 match the i-th start point condition acquired in step S403. If the one or more pieces of feature information match the i-th start point condition, the process proceeds to step S405; if not, the process proceeds to step S410.

[0161] (Step S405) The transmitter 132 acquires, from the start point storage unit 112, one or more node identifiers that are paired with the i-th start point condition.

[0162] (Step S406) The transmission unit 132 assigns 1 to the counter j.

[0163] (Step S407) The transmission unit 132 determines whether or not the j-th node identifier exists among the one or more node identifiers acquired in step S405. If the j-th node identifier exists, the process proceeds to step S408; if not, the process proceeds to step S410.

[0164] (Step S408) The transmission unit 132 performs an ignition process for the node identified by the j-th node identifier. An example of the ignition process will be described with reference to the flowchart in Fig. 5. The ignition process is a process that is performed when a node ignites.

[0165] (Step S409) The transmission unit 132 increments the counter j by 1. The process returns to step S407.

[0166] (Step S410) The transmission unit 132 increments the counter i by 1. The process returns to step S402.

[0167] 4, if there is no start point condition in the start point storage unit 112, the transmission unit 132 proceeds to step S406 after acquiring one or more node identifiers from the start point storage unit 112. In this case, the node that will fire first has already been determined.

[0168] Next, an example of the firing process in steps S408 and S612 will be described with reference to the flowchart of FIG.

[0169] (Step S501) The transmission unit 132 acquires the node identifier of the node that will fire. Note that this node identifier is, for example, the j-th node identifier of step S407.

[0170] (Step S502) The transmission unit 132 acquires one or more pieces of characteristic information and associates the one or more pieces of characteristic information with the node identifier acquired in step S501.

[0171] The one or more pieces of characteristic information are characteristic information acquired from the received information or characteristic information associated with the node from which the firing is transmitted. The characteristic information associated with the node from which the firing is transmitted is usually the same as the characteristic information acquired from the received information. The characteristic information associated with the node from which the firing is transmitted is characteristic information among the characteristic information acquired from the received information, and may be only the characteristic information used to determine the firing condition.

[0172] (Step S503) The transmission unit 132 acquires timer information that identifies the time of firing. The transmission unit 132 generates firing information that includes the node identifier acquired in step S501 and the timer information. The transmission unit 132 stores the firing information in the storage unit 11.

[0173] The transmission unit 132 may obtain, for example, timer information specifying the time of firing from a clock (not shown). The transmission unit 132 may obtain, for example, the node identifier of the node that caused the firing and the edge identifier of the edge that was passed through when the firing occurred, and may form firing information having the node identifier and the edge identifier, and store the firing information in the storage unit 11.

[0174] (Step S504) The transmission unit 132 updates one or more node attribute values ​​paired with the node identifier acquired in step S501, and returns to the upper-level processing.

[0175] Here, the node attribute values ​​updated by the transmission unit 132 are, for example, count information and firing probability information. For example, the transmission unit 132 increments the count information paired with the node identifier acquired in step S501 by 1. For example, the transmission unit 132 increments the firing probability information paired with the node identifier acquired in step S501. Note that the amount of increment is not important. For example, the transmission unit 132 substitutes the original firing probability information into an increasing function that uses the firing probability information as a parameter, executes the increasing function, acquires new firing probability information, and rewrites the original firing probability information with the new firing probability information.

[0176] Next, an example of the next transmission process in step S308 will be described with reference to the flowchart of FIG.

[0177] (Step S601) The transmission unit 132 acquires the node identifier of a node of interest. The node of interest is the i-th firing node of step S307.

[0178] (Step S602) The transmitter 132 acquires one or more pieces of feature information paired with the node identifier acquired in step S601. The one or more pieces of feature information paired with the node identifier are nodes connected by edges to the node identified by the node identifier, and may be one or more pieces of feature information paired with a node that fired previously.

[0179] (Step S603) The transmission unit 132 obtains one or more edge identifiers that are paired with the node identifier obtained in step S601 from the NN storage unit 111. Note that these one or more edge identifiers are identifiers of edges ahead of the node. The node ahead refers to the destination to which a signal is transmitted from the node. The transmission of a signal can be considered as the passing of one or more pieces of feature information.

[0180] (Step S604) The transmission unit 132 determines whether or not one or more edge identifiers were acquired in step S603. If an edge identifier was acquired, the process proceeds to step S605; if an edge identifier was not acquired, the process returns to the upper level processing. If an edge identifier paired with the node identifier was not acquired, this means that the node identified by the node identifier is a node in the output layer of the NN node group.

[0181] (Step S605) The transmission unit 132 assigns 1 to the counter i.

[0182] (Step S606) The transmission unit 132 determines whether or not the i-th edge identifier exists among the edge identifiers acquired in step S603. If the i-th edge identifier exists, the process proceeds to step S607; if not, the process returns to the upper level process.

[0183] (Step S607) The transmission unit 132 acquires one or more edge attribute values ​​paired with the i-th edge identifier from the NN storage unit 111. The one or more edge attribute values ​​include, for example, a weight.

[0184] (Step S608) The transmission unit 132 determines whether or not to transmit a signal via the edge, using one or more edge attribute values ​​(e.g., weights) acquired in step S607. If the signal is to be transmitted, the process proceeds to step S609; if not, the process proceeds to step S614.

[0185] Note that the transmission unit 132 normally increases the probability of determining to transmit a signal as the weight increases. The transmission unit 132 obtains the probability of determining to transmit a signal, for example, using an increasing function with the weight as a parameter, and determines whether to transmit a signal in accordance with the probability. The transmission unit 132 generates a random number, for example, and uses the random number to determine whether to transmit a signal in accordance with the probability.

[0186] (Step S609) The transmission unit 132 acquires node information of the node at the end of the edge. More specifically, the transmission unit 132 acquires the node identifier of the node to which the signal is to be transmitted, which is a node identifier that pairs with the i-th edge identifier, and acquires the node information of the node identified by the node identifier from the NN storage unit 111.

[0187] (Step S610) The transmission unit 132 performs firing determination processing for the node using the node information acquired in step S609. An example of the firing determination processing will be described with reference to the flowchart of FIG.

[0188] (Step S611) If the determination result in step S610 is "fire", the communication unit 132 proceeds to step S612, and if the determination result is "not fire", the communication unit 132 proceeds to step S614.

[0189] (Step S612) The transmission unit 132 performs an in-fire process for the node identified by the node information acquired in step S609. The in-fire process is the process shown in the flowchart of FIG.

[0190] (Step S613) The transmission unit 132 performs a next transmission process for the node identified by the node information acquired in step S609. The next transmission process is the process shown in the flowchart in FIG.

[0191] (Step S614) The transmission unit 132 increments the counter i by 1. The process returns to step S606.

[0192] Next, an example of the firing determination process in step S610 will be described with reference to the flowchart of FIG.

[0193] (Step S701) The transmission unit 132 acquires a firing condition corresponding to a node of interest from the firing condition storage unit 113. For example, the transmission unit 132 acquires from the firing condition storage unit 113 a firing condition paired with a node identifier included in the node information of the node of interest.

[0194] (Step S702) The transmission unit 132 acquires one or more pieces of feature information corresponding to the node of interest.

[0195] (Step S703) The transmission unit 132 determines whether or not the one or more pieces of feature information acquired in step S702 satisfy the firing condition acquired in step S701. If the firing condition is satisfied, the process proceeds to step S704, and if the firing condition is not satisfied, the process proceeds to step S705.

[0196] (Step S704) The communication unit 132 assigns "fire" to the variable "determination result" and returns to the upper level process.

[0197] (Step S705) The communication unit 132 assigns "not fired" to the variable "determination result" and returns to the upper level process.

[0198] In the flowchart of FIG. 7, the transmission unit 132 may determine whether or not to fire based on the probability according to the firing probability information included in the node information.

[0199] Next, an example of the survival rule application process in step S203 will be described with reference to the flowchart of FIG.

[0200] (Step S801) The determination unit 137 assigns 1 to a counter i.

[0201] (Step S802) The judgment unit 137 judges whether or not the i-th survival rule exists in the survival rule storage unit 117. If the i-th survival rule exists, the process proceeds to step S803; if not, the process returns to the upper process.

[0202] (Step S803) The determination unit 137 acquires the i-th survival rule from the survival rule storage unit 117.

[0203] (Step S804) The judgment unit 137 judges whether the i-th survival rule acquired in step S803 corresponds to the feature set acquired by the feature acquisition unit 131. If they correspond, the process proceeds to step S805, and if they do not correspond, the process proceeds to step S807. It is assumed that the i-th survival rule corresponds to the feature set acquired by the feature acquisition unit 131, which means that the i-th survival rule matches.

[0204] In this step, the determination unit 137 acquires the feature set acquired by the feature acquisition unit 131. This feature set is, for example, a feature set paired with the ignition node. This feature set may be a part of one or more pieces of feature information acquired by the feature acquisition unit 131. Next, the determination unit 137 acquires a feature set paired with the i-th survival rule. Next, the determination unit 137 determines whether the feature set acquired by the feature acquisition unit 131 and the feature set paired with the i-th survival rule satisfy a predetermined condition. If the predetermined condition is satisfied, the process proceeds to step S805; if not, the process proceeds to step S807. The predetermined condition is, for example, that the similarity between the two feature sets is equal to or greater than a threshold.

[0205] (Step S805) The rule application unit 138 acquires, from the survival rule storage unit 117, the judgment information paired with the i-th survival rule.

[0206] (Step S806) The rule application unit 138 performs edge weight change processing in accordance with the determination information. An example of such weight change processing will be described with reference to the flowchart of FIG.

[0207] (Step S807) The determination unit 137 increments the counter i by 1. The process returns to step S802.

[0208] Next, an example of the weight change process in step S806 will be described with reference to the flowchart in FIG.

[0209] (Step S901) The rule application unit 138 acquires the obtained firing pattern.

[0210] (Step S902) The rule application unit 138 determines whether the obtained judgment information is positive or negative. If it is positive information, the process proceeds to step S903, and if it is negative information, the process proceeds to step S912. Note that the obtained judgment information is the judgment information acquired in step S805.

[0211] (Step S903) The rule application unit 138 assigns 1 to a counter i.

[0212] (Step S904) The rule application unit 138 determines whether the i-th ignition node exists in the ignition pattern acquired in step S901. If the i-th ignition node exists, the process proceeds to step S905; if not, the process returns to the upper level process.

[0213] (Step S905) The rule application unit 138 acquires the area identifier paired with the i-th firing node from the NN storage unit 111. The area identifier paired with the i-th firing node is the area identifier paired with the node identifier of the i-th firing node.

[0214] (Step S906) The rule application unit 138 determines whether the area identifier acquired in step S905 is a positive area identifier. If it is a positive area identifier, the process proceeds to step S907, and if it is not a positive area identifier, the process proceeds to step S911.

[0215] (Step S907) The rule application unit 138 assigns 1 to a counter j.

[0216] (Step S908) The rule application unit 138 determines whether or not there is a jth edge connected to the ith firing node by referring to the NN storage unit 111. If there is a jth edge, the process proceeds to step S909; if there is no jth edge, the process proceeds to step S911.

[0217] (Step S909) The rule application unit 138 increases the weight of the edge information of the j-th edge. The amount of increase is, for example, "1," but is not critical. Increasing the weight means changing the attribute value "weight" of the edge information.

[0218] (Step S910) The rule application unit 138 increments the counter j by 1. The process returns to step S908.

[0219] (Step S911) The rule application unit 138 increments the counter i by 1. The process returns to step S904.

[0220] (Step S912) The rule application unit 138 assigns 1 to a counter i.

[0221] (Step S923) The rule application unit 138 determines whether the i-th ignition node exists in the ignition pattern acquired in step S901. If the i-th ignition node exists, the process proceeds to step S905; if not, the process returns to the upper level process.

[0222] (Step S914) The rule application unit 138 acquires from the NN storage unit 111 the area identifier paired with the i-th firing node.

[0223] (Step S915) The rule application unit 138 determines whether the area identifier acquired in step S905 is a negative area identifier. If it is a negative area identifier, the process proceeds to step S916, and if it is not a negative area identifier, the process proceeds to step S920.

[0224] (Step S916) The rule application unit 138 assigns 1 to a counter j.

[0225] (Step S917) The rule application unit 138 refers to the NN storage unit 111 and determines whether or not there is a jth edge connected to the ith firing node. If there is a jth edge, the process proceeds to step S918; if there is no jth edge, the process proceeds to step S920. Note that the edge connected to the firing node is identified by an edge identifier that pairs with the node identifier of the firing node.

[0226] (Step S918) The rule application unit 138 increases the weight of the edge information of the j-th edge. Note that the amount of increase is, for example, "1", but is not critical.

[0227] (Step S919) The rule application unit 138 increments the counter j by 1. The process returns to step S917.

[0228] (Step S920) The rule application unit 138 increments the counter i by 1. The process returns to step S913.

[0229] Next, an example of the firing pattern processing in step S205 will be described with reference to the flowchart of FIG.

[0230] (Step S1001) The corpus call OSAM unit 133 determines one or more actions to be executed according to the firing pattern, and acquires an action identifier that identifies each of the one or more actions. An example of such action determination processing will be described with reference to the flowchart in FIG. 11.

[0231] (Step S1002) The corpus call summing unit 133 assigns 1 to a counter i.

[0232] (Step S1003) The corpus call sum unit 133 determines whether the i-th action identifier exists among the one or more action identifiers acquired in step S1001. If the i-th action identifier exists, the process proceeds to step S1004; if not, the process returns to the upper level process.

[0233] (Step S1004) The corpus call summing unit 133 acquires the action information paired with the i-th action identifier from the action management unit 114.

[0234] (Step S1005) The corpus call OSAM unit 133 executes an operation using the operation information acquired in step S1004.

[0235] (Step S1006) The corpus call summing unit 133 performs post-processing. An example of the post-processing will be described with reference to the flowchart in Fig. 14. Note that the post-processing is processing performed after the execution of an operation.

[0236] (Step S1007) The corpus call summing unit 133 increments the counter i by 1. The process returns to step S1003.

[0237] Next, an example of the operation determination process in step S1001 will be described with reference to the flowchart in FIG.

[0238] (Step S1101) Corpus call OSAM unit 133 references cortex unit 115, which is a long-term memory, and acquires action identifiers of candidate actions to be executed. This process is called long-term memory reference processing. An example of long-term memory reference processing will be described with reference to the flowchart in FIG. 12.

[0239] (Step S1102) The corpus call SAM unit 133 references the hippocampus unit 116, which is a short-term memory, and acquires the action identifier of the action that is a candidate for execution. This process is called short-term memory reference processing. An example of short-term memory reference processing will be described with reference to the flowchart in FIG. 13.

[0240] (Step S1103) The action determination means 1332 acquires a score based on the expected value of the long-term memory information for each of the one or more pieces of long-term memory information temporarily accumulated in step S1101. The action determination means 1332 acquires a score based on the expected value of the short-term memory information for each of the one or more pieces of short-term memory information temporarily accumulated in step S1102. Next, the action determination means 1332 determines one or more scores that satisfy the action conditions from the acquired scores. Next, the action determination means 1332 acquires action identifiers contained in the long-term memory information or short-term memory information corresponding to each of the one or more determined scores.

[0241] The action determining means 1332 may obtain the expected value as the score, or may obtain the score using the expected value and the similarity between the two firing patterns.

[0242] The operating conditions are, for example, "the score is the highest," "the score is equal to or greater than the threshold," or "the score is in the top N (N is a natural number equal to or greater than 2)."

[0243] Next, an example of the long-term memory reference process in step S1101 will be described with reference to the flowchart in FIG.

[0244] (Step S1201) The pattern acquiring unit 1331 assigns 1 to a counter i.

[0245] (Step S1202) The pattern acquisition means 1331 determines whether or not the i-th ignition node exists. If the i-th ignition node exists, the process proceeds to step S1203; if not, the process proceeds to step S1207. Note that the information identifying the i-th ignition node is the i-th node information in the storage unit 11.

[0246] (Step S1203) The pattern acquiring unit 1331 acquires the area identifier corresponding to the i-th node information from the storage unit 11.

[0247] (Step S1204) The pattern acquisition unit 1331 determines whether the area identifier acquired in step S1203 is "Cortex." If the acquired area identifier is "Cortex," the process proceeds to step S1205, and if the acquired area identifier is not "Cortex," the process proceeds to step S1206.

[0248] (Step S1205) The pattern acquiring unit 1331 temporarily stores the i-th node information in a buffer (not shown) in the storage unit 11. The buffer stores node information including node identifiers that constitute the cortex firing pattern.

[0249] (Step S1206) The pattern acquiring unit 1331 increments the counter i by 1. The process returns to step S1202.

[0250] (Step S1207) The node identifiers included in the one or more pieces of node information stored in step S1205 are acquired, and a firing pattern including one or more node identifiers is constructed. The firing pattern here is called a cortex firing pattern.

[0251] (Step S1208) The action determining means 1332 assigns 1 to the counter j.

[0252] (Step S1209) Operation determining means 1332 determines whether or not the jth long-term memory information exists in cortex unit 115. If the jth long-term memory information exists, the process proceeds to step S1210; if not, the process returns to the upper level process.

[0253] (Step S1210) The action determining means 1332 acquires the time information contained in the j-th long-term memory information.

[0254] (Step S1211) The operation determination unit 1332 determines whether the time division obtained in step S1210 satisfies the first time condition. If the first time condition is satisfied, the process proceeds to step S1212. If the first time condition is not satisfied, the process proceeds to step S1216.

[0255] (Step S1212) The action determining means 1332 acquires the firing pattern contained in the j-th long-term memory information.

[0256] (Step S1213) The action determining unit 1332 obtains the similarity between the two firing patterns, that is, the cortex firing pattern obtained in step S1207 and the firing pattern obtained in step S1212.

[0257] (Step S1214) The action determining unit 1332 determines whether the similarity acquired in step S1213 satisfies the similarity condition. If the similarity condition is satisfied, the process proceeds to step S1215, and if the similarity condition is not satisfied, the process proceeds to step S1216.

[0258] (Step S1215) The action determining means 1332 accumulates the j-th long-term memory information in a buffer (not shown).

[0259] (Step S1216) The action determination means 1332 increments the counter j by 1. The process returns to step S1209.

[0260] In addition, this may be done in the cases of Figures 1 to 3.

[0261] Next, an example of the short-term memory reference process in step S1102 will be described with reference to the flowchart in FIG.

[0262] (Step S1301) The pattern acquiring unit 1331 assigns 1 to a counter i.

[0263] (Step S1302) The pattern acquisition means 1331 determines whether or not the i-th ignition node exists. If the i-th ignition node exists, the process proceeds to step S1303; if not, the process proceeds to step S1307. Note that the information identifying the i-th ignition node is the i-th node information in the storage unit 11.

[0264] (Step S 1303 ) The pattern acquiring unit 1331 acquires the area identifier corresponding to the i-th node information from the storage unit 11 .

[0265] (Step S1304) The pattern acquisition unit 1331 determines whether the area identifier acquired in step S1303 is "Hippocampus." If the acquired area identifier is "Hippocampus," the process proceeds to step S1305, and if the acquired area identifier is not "Hippocampus," the process proceeds to step S1306.

[0266] (Step S1305) The pattern acquiring unit 1331 temporarily stores the i-th node information in a buffer (not shown) in the storage unit 11. The buffer stores node information including node identifiers that constitute the hippocampus firing pattern.

[0267] (Step S1306) The pattern acquiring unit 1331 increments the counter i by 1. The process returns to step S1302.

[0268] (Step S1307) The node identifiers included in the one or more pieces of node information stored in step S1305 are acquired, and a firing pattern including the one or more node identifiers is constructed. This firing pattern is called a hippocampus firing pattern.

[0269] (Step S1308) The action determining means 1332 assigns 1 to the counter j.

[0270] (Step S1309) The operation determination means 1332 determines whether or not the jth short-term memory information exists in the hippocampus unit 116. If the jth short-term memory information exists, the process proceeds to step S1310; if not, the process returns to the upper level process.

[0271] (Step S1310) The action determining means 1332 acquires the time information contained in the j-th short-term memory information.

[0272] (Step S1311) The operation determination means 1332 determines whether or not the time division obtained in step S1310 satisfies the second time condition in the storage unit 11. If the second time condition is satisfied, the process proceeds to step S1312, and if the second time condition is not satisfied, the process proceeds to step S1316.

[0273] (Step S1312) The action determining means 1332 acquires the firing pattern contained in the j-th short-term memory information.

[0274] (Step S1313) The action determining unit 1332 obtains the similarity between the two firing patterns, that is, the cortex firing pattern obtained in step S1307 and the firing pattern obtained in step S1312.

[0275] (Step S1314) The action determining unit 1332 determines whether the similarity acquired in step S1313 satisfies the similarity condition. If the similarity condition is satisfied, the process proceeds to step S1315, and if the similarity condition is not satisfied, the process proceeds to step S1316.

[0276] (Step S1315) The action determination means 1332 accumulates the j-th short-term memory information in a buffer (not shown).

[0277] (Step S1316) The action determination means 1332 increments the counter j by 1. The process returns to step S1309.

[0278] Next, an example of the post-processing in step S1006 will be described with reference to the flowchart in FIG.

[0279] (Step S1401) The corpus call sum unit 133 stores an action identifier for identifying a performed action. The corpus call sum unit 133 usually stores the action identifier in the hippocampus unit 116.

[0280] (Step S1402) The information receiving unit 121 determines whether new information has been received after the operation. If new information has been received, the process proceeds to step S1404, and if not, the process proceeds to step S1403.

[0281] (Step S1403) The processing unit 13 determines whether a timeout has occurred. If a timeout has occurred, the processing returns to the upper processing, and if a timeout has not occurred, the processing returns to step S1402. Note that, for example, the processing unit 13 determines that a timeout has occurred if a predetermined time or more has elapsed since the execution of the operation.

[0282] (Step S1404) The transmission unit 132 etc. performs the ignition transmission process using the new information received in step S1402. An example of the ignition transmission process has been described with reference to the flowchart of FIG.

[0283] (Step S1405) The processing unit 13 performs a degree acquisition process. An example of the degree acquisition process will be described with reference to the flowchart in FIG.

[0284] (Step S1406) The corpus call summing unit 133 acquires a reward using the positivity and negativity acquired in step S1405. The corpus call summing unit 133 acquires the reward in combination with the action identifier of the performed action.

[0285] (Step S1407) The corpus call OSAM unit 133 uses the reward acquired in step S1406 to update the expected value paired with the action identifier of the performed action. Note that updating the expected value may involve adding a new expected value. For example, the corpus call OSAM unit 133 adds the acquired reward to the expected value paired with the action identifier. Here, it is preferable that the corpus call OSAM unit 133 also updates the feature set paired with the action identifier of the performed action. For example, the corpus call OSAM unit 133 updates the feature set corresponding to the action identifier using the feature set acquired by the feature acquisition unit 131. Furthermore, the corpus call OSAM unit 133 may associate a pair of a new expected value and the feature set acquired by the feature acquisition unit 131 with the action identifier of the performed action.

[0286] Next, an example of the degree acquisition process in step S1405 will be described with reference to the flowchart in FIG.

[0287] (Step S1501) The processing unit 13 acquires the obtained firing pattern. Here, the processing unit 13 may acquire the node identifier included in each piece of firing information stored in the storage unit 11.

[0288] (Step S1502) The processing unit 13 initializes the positivity and negativity. Note that the initialization usually involves setting the positivity to "0" and the negativity to "0", but this is not limiting.

[0289] (Step S1503) The processing unit 13 assigns 1 to the counter i.

[0290] (Step S1504) The processing unit 13 determines whether or not the i-th ignition node exists in the ignition pattern acquired in step S1501. If the i-th ignition node exists, the process proceeds to step S1505; if not, the process returns to the upper level process. Note that the i-th ignition node exists when, for example, the i-th ignition node identifier exists in the ignition pattern.

[0291] (Step S1505) The processing unit 13 acquires a node identifier for identifying the i-th firing node.

[0292] (Step S1506) The processing unit 13 acquires from the NN storage unit 111 an area identifier that is paired with the node identifier acquired in step S1505.

[0293] (Step S1507) The processing unit 13 determines whether the area identifier acquired in step S1506 is "Posia My Gudara." If it is "Posia My Gudara," the processing unit 13 proceeds to step S1508, and if it is not "Posia My Gudara," the processing unit 13 proceeds to step S1509.

[0294] (Step S1508) The positive feedback unit 134 increases the positivity level. For example, the positive feedback unit 134 increases the positivity level by a predetermined amount (for example, "1").

[0295] (Step S1509) The processing unit 13 determines whether the area identifier acquired in step S1506 is "Negaamigudara." If it is "Negaamigudara," the processing proceeds to step S1510, and if it is not "Negaamigudara," the processing proceeds to step S1511.

[0296] (Step S1510) The negativity unit 135 increases the negativity by, for example, a predetermined amount (for example, "1").

[0297] (Step S1511) The processing unit 13 increments the counter i by 1. The process returns to step S1504.

[0298] Next, an example of the time condition change process in step S207 will be described with reference to the flowchart of FIG.

[0299] (Step S1601) The time condition changing unit 136 acquires, for example, the total number of firing nodes, the number of cortex firing nodes, and the number of hippocampus firing nodes.

[0300] The total number of firing nodes is the number of all nodes currently firing. The number of Cortex firing nodes is the number of nodes that correspond to the area identifier "Cortex" among all firing nodes. The number of Hippocampus firing nodes is the number of nodes that correspond to the area identifier "Hippocampus" among all firing nodes.

[0301] The time condition changing unit 136 acquires, for example, the total number of ignition nodes, which is the number of all ignition information in the storage unit 11. The time condition changing unit 136 acquires, for example, the total number of ignition nodes, which is the number of node identifiers included in the acquired ignition patterns.

[0302] The time condition change unit 136, for example, obtains the number of Cortex firing nodes, which is the number of firing information having the area identifier "Cortex", and the number of Hippocampus firing nodes, which is the number of firing information having the area identifier "Hippocampus", from all the firing information in the storage unit 11.

[0303] The time condition change unit 136 acquires, for example, the number of cortex firing nodes, which is the number of area identifiers "cortex" that pair with the node identifiers of the acquired firing pattern, and the number of hippocampus firing nodes, which is the number of area identifiers "hippocampus" that pair with the node identifiers of the acquired firing pattern.

[0304] (Step S1602) The time condition change unit 136 uses the number of ignition nodes acquired in step S1601 to determine whether the first change condition is satisfied. If the first change condition is satisfied, the process proceeds to step S1603; if not, the process proceeds to step S1604. The number of ignition nodes acquired in step S1601 here is the number of cortex ignition nodes or the total number of ignition nodes.

[0305] (Step S1603) The time condition change unit 136 obtains a first time condition to be updated by using the number of ignition nodes used to determine whether the first change condition is satisfied. The time condition change unit 136 overwrites the original first time condition with the first time condition.

[0306] (Step S1604) The time condition change unit 136 uses the number of ignition nodes acquired in step S1601 to determine whether the second change condition is satisfied. If the second change condition is satisfied, the process proceeds to step S1605; if not, the process returns to the upper level process. The number of ignition nodes acquired in step S1601 here is the number of hippocampus ignition nodes or the total number of ignition nodes.

[0307] (Step S1605) The time condition change unit 136 acquires a second time condition to be updated using the number of ignition nodes used to determine whether the second change condition is satisfied. The time condition change unit 136 overwrites the original second time condition with the second time condition. The process returns to the upper level process.

[0308] As described above, according to this embodiment, it is possible to simulate brain activity when performing actions that take into account long-term memory and short-term memory, thereby providing an information processing device 1 that can behave like an adult or like a child.

[0309] Furthermore, according to this embodiment, the expected value when performing an operation can be updated.

[0310] Furthermore, according to this embodiment, it is possible to store short-term memory information including the action identifier of the executed action.

[0311] Furthermore, according to this embodiment, it is possible to store long-term memory information including the action identifier of the action that has been performed.

[0312] Furthermore, according to this embodiment, an appropriate neural network can be constructed using managed survival rules.

[0313] The processing in this embodiment may be implemented by software. This software may be distributed by software download or the like. Furthermore, this software may be recorded on a recording medium such as a CD-ROM and distributed. This also applies to other embodiments in this specification. The software that implements the information processing device 1 in this embodiment is the following program. That is, this program includes: an NN storage unit that stores neural network information having two or more pieces of node information each having a node identifier that identifies a node and corresponding to an area identifier that identifies the area of ​​the node; and one or more pieces of edge information each having an edge identifier that identifies an edge and specifying a connection between the nodes; a start point storage unit that stores one or more pieces of start point information each having one or more node identifiers that identify nodes that will fire without passing through other nodes, the start point condition being a condition related to a feature set that is one or more pieces of feature information of the received information, and being a condition for determining a node that will fire without passing through other nodes; a firing condition storage unit that stores firing conditions related to the feature set, the firing conditions being firing conditions associated with one or more pieces of node information among the two or more pieces of node information; an action management unit that stores action information each corresponding to two or more action identifiers that identify actions; a cortex unit storing one or more pieces of long-term memory information having a firing pattern having one or more firing node identifiers for identifying a node, a past action identifier for identifying an action performed in the past, and an expected value of the action; and a hippocampus unit storing one or more pieces of short-term memory information having a firing pattern, one or more short-term action identifiers for identifying an action performed within a predetermined period, and an expected value of the action, the computer comprising: a reception unit for receiving information; a feature acquisition unit for acquiring a feature set, which is one or more pieces of feature information, from the information received by the reception unit; and a node identifier that is paired with a start point condition that matches the feature set acquired by the feature acquisition unit, the node identifier being connected by an edge to the fired node identified by each of the one or more node identifiers, and to which all or a part of the feature set is passed;a communication unit that performs a communication process to determine a firing node identifier of a node that fires based on a firing condition corresponding to the node; a program that acquires long-term firing patterns having one or more firing node identifiers that are paired with an area identifier indicating a cortex region, the long-term firing patterns being one or more firing node identifiers determined by the communication unit; acquires scores based on expected values ​​that are paired with the firing patterns stored in the cortex unit, the short-term firing patterns having one or more firing node identifiers that are paired with an area identifier indicating a hippocampus region, the short-term firing patterns being one or more firing node identifiers determined by the communication unit; acquires scores based on expected values ​​that are paired with the firing patterns stored in the hippocampus region, the short-term firing patterns being one or more firing node identifiers that are paired with an area identifier indicating a hippocampus region, the short-term firing patterns being firing patterns that are similar to the short-term firing patterns, the short-term firing patterns being firing patterns that satisfy a similarity condition, the short-term firing patterns being firing patterns

[0314] FIG. 17 is a block diagram of a computer system 300 that executes the programs described in this specification to realize the information processing device 1 according to the various embodiments described above.

[0315] In FIG. 17, a computer system 300 includes a computer 301 including a CD-ROM drive, a keyboard 302, a mouse 303, and a monitor 304.

[0316] 17, the computer 301 includes, in addition to a CD-ROM drive 3012, an MPU 3013, a bus 3014 connected to the CD-ROM drive 3012 etc., a ROM 3015 for storing programs such as a boot-up program, a RAM 3016 connected to the MPU 3013 for temporarily storing instructions for application programs and providing temporary storage space, and a hard disk 3017 for storing application programs, system programs, and data. Although not shown here, the computer 301 may further include a network card for providing connection to a LAN.

[0317] A program that causes computer system 300 to execute the functions of information processing device 1 of the above-described embodiment may be stored on CD-ROM 3101, inserted into CD-ROM drive 3012, and then transferred to hard disk 3017. Alternatively, the program may be transmitted to computer 301 via a network (not shown) and stored on hard disk 3017. The program is loaded into RAM 3016 when executed. The program may also be loaded directly from CD-ROM 3101 or the network.

[0318] The program does not necessarily have to include an operating system (OS) or a third-party program that causes the computer 301 to execute the functions of the information processing device 1 of the above-described embodiment. The program only needs to include instructions that call appropriate functions (modules) in a controlled manner to achieve the desired results. How the computer system 300 operates is well known, and a detailed description thereof will be omitted.

[0319] The computer that executes the program may be a single computer or a plurality of computers, that is, it may perform centralized processing or distributed processing.

[0320] Furthermore, in each of the above embodiments, each process may be realized by centralized processing in a single device, or may be realized by distributed processing in a plurality of devices.

[0321] The present invention is not limited to the above-described embodiment, and various modifications are possible, and it goes without saying that these modifications are also included within the scope of the present invention.

[0322] As described above, the information processing device 1 according to the present invention has the effect of being able to simulate the activity in the brain when performing an action taking into account long-term memory and short-term memory, and is useful as a robot or the like.

Claims

1. An NN storage unit that stores neural network information having two or more node information pieces each having a node identifier that identifies a node and corresponding to an area identifier that identifies the area of ​​the node, and one or more edge information pieces each having an edge identifier that identifies an edge and specifying a connection between nodes; a starting point storage unit that stores one or more starting point information pieces each having one or more node identifiers that identify a node that will fire without passing through other nodes, which are conditions related to a feature set that is one or more feature information pieces of received information and are conditions for determining a node that will fire without passing through other nodes; a firing condition storage unit that stores firing conditions related to the feature set, which are firing conditions related to one or more node information pieces of the two or more node information pieces; an action management unit that stores action information pieces each corresponding to two or more action identifiers that identify an action; a cortex unit that stores one or more long-term memory information pieces each having a firing pattern having one or more firing node identifiers that identify a node to fire, a past action identifier that identifies an action that was performed in the past, and an expected value of the action; a hippocampus unit that stores one or more pieces of short-term memory information having one or more short-term action identifiers that identify an action performed during a firing pattern and a predetermined period of time, and an expected value of the action; a reception unit that receives information; a feature acquisition unit that acquires a feature set that is one or more pieces of feature information from the information received by the reception unit; and a transmission unit that determines, from the start point storage unit, one or more node identifiers that are paired with a start point condition that matches the feature set acquired by the feature acquisition unit, and performs a transmission process that determines the firing node identifier of the node that is connected by an edge to the fired node identified by the one or more node identifiers, is a node to which all or a part of the feature set is passed, and that will fire based on the firing condition corresponding to the node.and a corpus call OSAM unit that acquires long-term firing patterns, each having one or more firing node identifiers determined by the transmission unit and paired with an area identifier indicating a cortex region, acquires a score based on an expected value that is a firing pattern that satisfies a similarity condition for the long-term firing pattern and is paired with the firing pattern stored in the cortex unit, acquires short-term firing patterns, each having one or more firing node identifiers determined by the transmission unit and paired with an area identifier indicating a hippocampus region, acquires a score based on an expected value that is a firing pattern that satisfies a similarity condition for the short-term firing pattern and is paired with the firing pattern stored in the hippocampus region, acquires an action identifier that is a past action identifier or a short-term action identifier for a score that satisfies an action condition from the two or more acquired scores, acquires action information that is paired with the action identifier from the action management unit, and performs an action according to the action information.

2. The information processing device of claim 1, further comprising a storage unit in which a first time condition, which is a condition relating to the time information and is a condition relating to the time when the long-term memory information is applied, is stored, and the corpus call analysis unit acquires, from among the long-term firing patterns paired with the time information that satisfies the first time condition, one or more firing node identifiers determined by the transmission unit and paired with an area identifier indicating a cortex region.

3. The information processing device according to claim 2, further comprising a condition change unit that acquires the number of ignition node identifiers determined by said transmission unit and changes said temporary condition using said number.

4. An information processing device as described in any one of claims 1 to 3, wherein the short-term memory information has time information regarding when the information was accumulated, a second time condition is stored which is a condition for applying the short-term memory information and is a condition regarding the time information, and the corpus call OSAM unit acquires, from among the short-term firing patterns paired with time information that satisfies the second time condition, one or more firing node identifiers determined by the transmission unit and which are paired with an area identifier indicating a hippocampus region.

5. The information processing device according to claim 4, further comprising a time condition changing unit that acquires the number of ignition node identifiers determined by said transmission unit and changes said second time condition using said number.

6. The information processing device described in any one of claims 1 to 5, wherein the reception unit further comprises: a positive amygdala unit that receives new information after the corpus call summing unit has performed an action corresponding to the action information, and uses the firing pattern when the reception unit receives the new information to obtain the number of fired nodes out of one or more nodes corresponding to an area identifier indicating a positive amygdala area, and obtains a positivity level using the number of nodes; and a negative amygdala unit that uses the firing pattern when the reception unit receives the new information to obtain the number of fired nodes out of one or more nodes corresponding to an area identifier indicating a negative amygdala area, and obtains a negativity level using the number of nodes; and the corpus call summing unit obtains a reward related to the difference between the positivity level obtained by the positive amygdala unit and the negativity level obtained by the negativity unit, and uses the reward to update an expectation value corresponding to an action identifier that identifies the action performed by the corpus call summing unit.

7. An information processing device as described in any one of claims 1 to 6, wherein the corpus call OSAM unit stores in the hippocampus unit short-term memory information having the action identifier that identifies the action performed by the corpus call OSAM unit and the firing pattern when the action is performed.

8. An information processing device as described in any one of claims 1 to 7, wherein the corpus call OSAM unit updates the frequency of the action in association with the action identifier that identifies the action performed by the corpus call OSAM unit, and when the frequency satisfies a long-term memory condition, the corpus call OSAM unit accumulates long-term memory information having the action identifier that identifies the action performed by the corpus call OSAM unit and the firing pattern when the action is performed in the cortex unit, and deletes short-term memory information including the action identifier from the hippocampus unit.

9. The edge information has edge weights, and the transmission unit is a node that is connected to the fired node by an edge, and when determining a node identifier of the node that will fire based on a firing condition corresponding to the node, performs processing such that the greater the weight of the edge, the more likely the node is to be determined as the node that will fire; a survival rule storage unit that stores one or more survival rules that are positive rules corresponding to positive information indicating a positive state or negative rules corresponding to negative information indicating a negative state and that correspond to a feature set; and a judgment unit that determines a survival rule that corresponds to the feature set acquired by the feature acquisition unit and acquires judgment information that is positive information or negative information associated with the survival rule.

7. The information processing device of claim 6, further comprising a rule application unit that, when the judgment information is positive information, increases the weight of the edge connecting the node that is the node ahead of the fired node and corresponds to the area identifier of the positive area, and, when the judgment information is negative information, increases the weight of the edge connecting the node that is the node ahead of the fired node and corresponds to the area identifier of the negative area.

10. An NN storage unit storing neural network information having a node identifier for identifying a node and two or more pieces of node information corresponding to an area identifier for identifying the area of ​​the node, and having an edge identifier for identifying an edge and one or more pieces of edge information specifying connections between nodes; a starting point storage unit storing one or more pieces of starting point information having one or more node identifiers for identifying nodes that will fire without passing through other nodes, in association with a starting point condition that is a condition related to a feature set that is one or more pieces of feature information of received information and is a condition for determining a node that will fire without passing through other nodes; and a firing condition corresponding to each of one or more pieces of node information among the two or more pieces of node information. an action management unit storing action information corresponding to two or more action identifiers for identifying actions; a cortex unit storing one or more pieces of long-term memory information each having a firing pattern with one or more firing node identifiers for identifying a firing node, a past action identifier for identifying an action performed in the past, and an expected value of the action; a hippocampus unit storing one or more pieces of short-term memory information each having a firing pattern, one or more short-term action identifiers for identifying actions performed within a predetermined period, and an expected value of the action; a reception unit; a feature acquisition unit; a transmission unit; and a corpus callosum unit, a receiving step in which the receiving unit receives information; a feature acquisition step in which the feature acquisition unit acquires a feature set, which is one or more pieces of feature information, from the information received by the receiving unit; and a transmission step in which the transmission unit determines, from the start point storage unit, one or more node identifiers that form a pair with a start point condition that matches the feature set acquired by the feature acquisition unit, and performs a transmission process in which the transmission unit determines an ignition node identifier of a node that is connected by an edge to an ignition node identified by the one or more node identifiers, is a node to which all or a part of the feature set is passed, and that ignites based on the ignition condition corresponding to the node.and a corpus call OSAM step in which the corpus call OSAM unit acquires long-term firing patterns, which are one or more firing node identifiers determined by the transmission unit and have one or more firing node identifiers that pair with area identifiers indicating cortex regions, acquires scores based on expected values ​​that are firing patterns that satisfy a similarity condition for the long-term firing patterns and are paired with the firing patterns stored in the cortex unit, acquires short-term firing patterns, which are one or more firing node identifiers determined by the transmission unit and have one or more firing node identifiers that pair with area identifiers indicating hippocampus regions, acquires scores based on expected values ​​that are firing patterns that satisfy a similarity condition for the short-term firing patterns and are paired with the firing patterns stored in the hippocampus unit, acquires action identifiers that are past action identifiers or short-term action identifiers for scores that satisfy an action condition from the two or more acquired scores, acquires action information that pairs with the action identifiers from the action management unit, and performs an action according to the action information.

11. An NN storage unit storing neural network information having a node identifier for identifying a node and two or more pieces of node information corresponding to an area identifier for identifying the area of ​​the node, and having an edge identifier for identifying an edge and one or more pieces of edge information specifying connections between nodes; a starting point storage unit storing one or more pieces of starting point information having one or more node identifiers for identifying nodes that fire without passing through other nodes, which is a condition related to a feature set that is one or more pieces of feature information of received information and is a condition for determining a node that fires without passing through other nodes; a computer that can access a firing condition storage unit that stores firing conditions related to a feature set, the firing conditions being associated with each node information; an action management unit that stores action information associated with two or more action identifiers that identify actions; a cortex unit that stores one or more long-term memory information pieces that have firing patterns having one or more firing node identifiers that identify firing nodes, past action identifiers that identify actions that were performed in the past, and expected values ​​of the actions; and a hippocampus unit that stores one or more short-term memory information pieces that have firing patterns, one or more short-term action identifiers that identify actions that were performed within a predetermined period, and expected values ​​of the actions, a reception unit that receives information; a feature acquisition unit that acquires a feature set, which is one or more pieces of feature information, from the information received by the reception unit; and a transmission unit that determines, from the start point storage unit, one or more node identifiers that form a pair with a start point condition that matches the feature set acquired by the feature acquisition unit, and performs a transmission process that determines an ignition node identifier of a node that is connected by an edge to an ignition node identified by the one or more node identifiers, is a node to which all or a part of the feature set is passed, and that ignites based on the ignition condition corresponding to the node.a program for acquiring a long-term firing pattern having one or more firing node identifiers determined by the transmission unit and paired with an area identifier indicating a cortex region, acquiring a score based on an expected value paired with the firing pattern stored in the cortex unit, which is a firing pattern that satisfies a similarity condition for the long-term firing pattern, and acquiring a short-term firing pattern having one or more firing node identifiers determined by the transmission unit and paired with an area identifier indicating a hippocampus region, which is a firing pattern that satisfies a similarity condition for the short-term firing pattern and paired with the firing pattern stored in the hippocampus region, acquiring an action identifier that is a past action identifier or a short-term action identifier for a score that satisfies an action condition from the two or more acquired scores, acquiring action information that pairs with the action identifier from the action management unit, and functioning as a corpus call OSAM unit that performs an action according to the action information.

Images