Apparatus and method for operating a multi-compartmental neuronal model, computer readable medium

Abstract

This disclosure provides an apparatus for computing a multi-compartment neuron model, comprising a coupling unit and a plurality of computing units connected to the coupling unit; wherein each computing unit is used to compute the initial membrane potential of each compartment as a virtual neuron; multiple compartments of the same multi-compartment neuron model correspond to different computing units; the coupling unit is used to acquire the initial membrane potentials obtained by each computing unit, compute the final membrane potentials of each compartment based on the initial membrane potentials of each compartment of the same multi-compartment neuron model, and send the final membrane potentials to the computing units corresponding to each compartment. This disclosure also provides a method, electronic device, and computer-readable medium for computing a multi-compartment neuron model.

Description

Technical Field

[0001] This disclosure relates to the field of neural network technology, and in particular to apparatus and methods for operating multi-compartment neuron models, and computer-readable media. Background Technology

[0002] Spiking Neural Networks (SNNs) and other neural networks have been widely used in fields such as image processing and speech recognition. A neural network can include multiple interconnected neuron models. Each neuron model can receive pulses from other neuron models and emit pulses to other neuron models when its own membrane potential exceeds a firing threshold.

[0003] To make the neuron model more accurately simulate biological neurons, a "multi-compartment neuron model" can be used, that is, referring to... Figure 1 A neuron model is divided into multiple "compartments" to obtain a multi-compartment neuron model composed of multiple compartments, where each compartment has its own membrane potential and the membrane potentials of different compartments can influence each other.

[0004] Therefore, efficient and accurate computation of multi-compartment neuron models is desired. Summary of the Invention

[0005] This disclosure provides an apparatus and method for operating a multi-compartment neuron model, as well as a computer-readable medium.

[0006] In a first aspect, embodiments of this disclosure provide an apparatus for computing multi-compartment neuron models, each of which includes multiple compartments; the apparatus includes a coupling unit and multiple computing units connected to the coupling unit; wherein...

[0007] Each of the aforementioned computational units is used to obtain the initial membrane potential of the commissure by using each corresponding commissure as a virtual neuron; multiple commissures of the same multi-commissure neuron model correspond to different computational units;

[0008] The coupling unit is used to obtain the initial membrane potential obtained by each computing unit, calculate the final membrane potential of each chamber based on the initial membrane potential of each chamber in the same multi-compartment neuron model, and send the final membrane potential to the computing unit corresponding to each chamber.

[0009] In some embodiments, at least the arithmetic unit corresponding to a compartment having dendrites has an input terminal.

[0010] In some embodiments, at least the arithmetic unit corresponding to a compartment having an axon has an output terminal.

[0011] Secondly, embodiments of this disclosure provide a method for operating a multi-compartment neuron model, performed using any of the apparatuses for operating a multi-compartment neuron model according to embodiments of this disclosure, the method comprising:

[0012] The computational unit uses each corresponding compartment as a virtual neuron to calculate the initial membrane potential of the compartment and sends the initial membrane potential to the coupling unit;

[0013] The coupling unit calculates the final membrane potential of each chamber based on the initial membrane potential of each chamber in the same multi-compartment neuron model, and sends the final membrane potential to the corresponding computation unit of each chamber.

[0014] In some embodiments, each of the computing units corresponds to only one room.

[0015] In some embodiments, the method of this disclosure is used to perform computation on multiple structurally identical multi-compartment neuron models;

[0016] Each of the aforementioned operational units corresponds to multiple compartments, and the multiple compartments corresponding to the same operational unit are the compartments in the same relative position in each multi-compartment neuron model.

[0017] In some embodiments, the computational unit uses each corresponding ventricle as a virtual neuron to calculate the initial membrane potential of the ventricle, including:

[0018] The computational unit uses multidimensional vector operations to obtain the initial membrane potential of each chamber as a virtual neuron; wherein, in the multidimensional vector operation, each chamber of the computational unit corresponds to one dimension of the multidimensional vector.

[0019] In some embodiments, at least the arithmetic unit corresponding to a compartment having dendrites has an input terminal;

[0020] The computational unit calculates the initial membrane potential of each ventricle using each corresponding ventricle as a virtual neuron, including: when the computational unit corresponding to a ventricle with dendrites receives an input signal from its input terminal, it determines that the virtual neuron of the corresponding ventricle has received an input event, and calculates the initial membrane potential of the ventricle.

[0021] In some embodiments, at least the arithmetic unit corresponding to a compartment having an axon has an output terminal;

[0022] The computational unit calculates the initial membrane potential of each ventricle as a virtual neuron, including: when it is determined that the virtual neuron corresponding to the ventricle with an axon fires, the computational unit corresponding to the ventricle sends an output signal to its output terminal.

[0023] Thirdly, embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements a method for computation of any of the multi-compartment neuron models of this disclosure.

[0024] In this embodiment of the present disclosure, the initial membrane potential of each commissure of the multi-commissure neuron model can be calculated in multiple computing units, with each commissure as a virtual neuron. Then, in the coupling unit, the initial membrane potentials of each commissure are coupled to obtain the final membrane potential of each commissure. Thus, this embodiment of the present disclosure can efficiently and accurately realize the operation of the multi-commissure neuron model through simple devices and calculations. Attached Figure Description

[0025] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the detailed embodiments to explain the present disclosure and do not constitute a limitation thereof. The above and other features and advantages will become more apparent to those skilled in the art from the description of the detailed embodiments with reference to the accompanying drawings, in which:

[0026] Figure 1 This is a schematic diagram of the structure of a multi-compartment neuron model;

[0027] Figure 2 A block diagram of a device for computing a multi-compartment neuron model provided in this embodiment of the disclosure;

[0028] Figure 3 A flowchart illustrating a method for computation of a multi-compartment neuron model provided in this disclosure embodiment;

[0029] Figure 4 A schematic diagram illustrating the grouping of compartments in multiple multi-compartment neuron models in the method for computing multi-compartment neuron models provided in this embodiment of the disclosure;

[0030] Figure 5 This is a schematic diagram of another multi-compartment neuron model;

[0031] Figure 6 This is a block diagram illustrating the composition of a computer-readable medium provided in an embodiment of the present disclosure. Detailed Implementation

[0032] To enable those skilled in the art to better understand the technical solutions of this disclosure, the apparatus and method for multi-compartment neuron model operation and the computer-readable medium provided in this disclosure are described in detail below with reference to the accompanying drawings.

[0033] The present disclosure will be described more fully below with reference to the accompanying drawings; however, the embodiments shown may be embodied in different forms and should not be construed as being limited to the embodiments set forth in this disclosure. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will enable those skilled in the art to fully understand the scope of this disclosure.

[0034] This disclosure may be described with reference to plan and / or cross-sectional views using the ideal schematic diagrams of this disclosure. Therefore, the example illustrations may be modified according to manufacturing techniques and / or tolerances.

[0035] Where there is no conflict, the various embodiments of this disclosure and the features thereof in the embodiments may be combined with each other.

[0036] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the disclosure. The term "and / or" as used in this disclosure includes any and all combinations of one or more of the associated enumerated entries. The singular forms "a" and "the" as used in this disclosure are also intended to include the plural forms, unless the context clearly indicates otherwise. The terms "comprising," "made of," etc., as used in this disclosure specify the presence of the stated feature, integral, step, operation, element, and / or component, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof.

[0037] Unless otherwise specified, all terms used in this disclosure (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this disclosure, and will not be interpreted as having an idealized or overly formal meaning, unless expressly so defined in this disclosure.

[0038] This disclosure is not limited to the embodiments shown in the accompanying drawings, but includes modifications to the configuration based on the manufacturing process. Therefore, the areas illustrated in the drawings are schematic, and the shapes of the areas shown illustrate specific shapes of the areas of an element, but are not intended to be limiting.

[0039] In a first aspect, embodiments of this disclosure provide an apparatus for computing a multi-compartment neuron model.

[0040] The multi-compartment neuron model of this disclosure is used for computation of the multi-compartment neuron model, with reference to... Figure 1 , Figure 5 Each multi-compartment neuron model includes multiple compartments.

[0041] Reference Figure 2 The apparatus for multi-compartment neuron model computation according to embodiments of the present disclosure includes a coupling unit and a plurality of computation units connected to the coupling unit.

[0042] Each computational unit is used to obtain the initial membrane potential of the ventricle by operating on each of its corresponding ventricles as a virtual neuron; multiple ventricles of the same multi-ventricle neuron model correspond to different computational units.

[0043] The coupling unit is used to obtain the initial membrane potential obtained by each operation unit, calculate the final membrane potential of each chamber based on the initial membrane potential of each chamber in the same multi-chamber neuron model, and send the final membrane potential to the operation unit corresponding to each chamber.

[0044] This disclosure includes multiple computing units, each of which is a relatively independent physical device with computing capabilities, such as a device for processing conventional neuron models (e.g., a single neuron computing device).

[0045] Furthermore, each processing unit corresponds to one or more chambers, and each chamber can be processed "as" a "neuron (virtual neuron)." That is, based on the input of the virtual neuron in the current time step (such as whether it fires or receives a pulse), and the parameters of the virtual neuron (such as leakage, firing threshold, resting potential, etc.), the membrane potential of the virtual neuron is calculated, which is to say, the initial membrane potential of the chamber is calculated. In other words, the "initial membrane potential" refers to the membrane potential that the chamber should have in the current time step if each chamber is an independent neuron (the potentials between chambers do not affect each other).

[0046] Among them, "time beat" is the minimum time period for the multi-compartment neuron model to run, that is, the membrane potential of each compartment of the multi-compartment neuron model is updated once in each time beat.

[0047] Obviously, since the multiple chambers of the same multi-compartment neuron model correspond to different computational units, the initial membrane potential of each chamber of each multi-compartment neuron model can be obtained by the "synchronous" computation of their corresponding multiple computational units in a time step.

[0048] In a multi-compartment neuron model, the compartments are not independent neurons because their potentials influence each other within a time beat, for example, as shown in the reference model. Figure 1 The membrane potential of ventricle C1 affects the membrane potential of ventricle C2, and the membrane potential of ventricle C2 also affects the membrane potential of ventricle C1. Moreover, this effect does not occur in the form of "pulses" but in the form of "analog quantities". This effect is called the "coupling" of ventricle membrane potentials.

[0049] The embodiments of this disclosure also include a coupling unit connected to each operation unit. Therefore, in each time step, the coupling unit can obtain the initial membrane potential of each chamber of a multi-compartment neuron model obtained by each operation unit, and calculate the final membrane potential that each chamber should have after coupling based on the initial membrane potential. In other words, it calculates the final membrane potential that each chamber of the multi-compartment neuron model should have in this time step, thereby completing the operation of the multi-compartment neuron model.

[0050] In this embodiment of the present disclosure, the initial membrane potential of each commissure of the multi-commissure neuron model can be calculated in multiple computing units, with each commissure as a virtual neuron. Then, in the coupling unit, the initial membrane potentials of each commissure are coupled to obtain the final membrane potential of each commissure. Thus, this embodiment of the present disclosure can efficiently and accurately realize the operation of the multi-commissure neuron model through simple devices and calculations.

[0051] In some embodiments, at least the arithmetic unit corresponding to a compartment having dendrites has an input terminal.

[0052] In some embodiments, at least the arithmetic unit corresponding to a compartment having an axon has an output terminal.

[0053] Reference Figure 1 Each neuron model has one or more dendrites for input and one axon for output. Each axon can be connected to the dendrites of one or more other neurons via synapses.

[0054] In a multi-compartment neuron model, since the neuron model is divided into different compartments, the dendrites and axons are actually located in different compartments. Therefore, the corresponding inputs and outputs also correspond to different compartments. In other words, the input event (such as a pulse) is actually input into a certain compartment, and the output (such as a pulse) is actually generated from a certain compartment.

[0055] Since each chamber corresponds to a processing unit, if a processing unit corresponds to a chamber with dendrites, the processing unit needs to have an input terminal to obtain an input signal from the input terminal, indicating that its corresponding chamber has an input; while if a processing unit corresponds to a chamber with axons, the processing unit needs to have an output terminal to send an output signal from the output terminal when its corresponding chamber needs to release.

[0056] Of course, some compartments may correspond only to the cell body (Soma), and therefore have neither dendrites nor axons. In such cases, the compartments will not have direct input or output, and their membrane potential will be determined only by leakage, coupling, etc.

[0057] This could involve setting up input and output terminals for each processing unit based on the conditions of its corresponding room, but referring to... Figure 2 To ensure structural uniformity and improve versatility, all operational units can have the same structure, with input and output terminals. However, if the chamber corresponding to an operational unit does not include dendrites or axons, then the input / output terminal of the corresponding operational unit can be "invalid," meaning it is not connected to other structures.

[0058] Of course, when the device for multi-compartment neuron model operation processes multiple multi-compartment neuron models with interconnected relationships, the input and output terminals of each operation unit should be interconnected according to the intercompartmental connections of different multi-compartment neuron models.

[0059] Alternatively, if the device for processing multi-compartment neuron models only handles one multi-compartment neuron model, then the input and output terminals of devices for processing different multi-compartment neuron models can be interconnected.

[0060] Secondly, embodiments of this disclosure provide a method for operating a multi-compartment neuron model, which is performed based on any of the apparatuses for operating a multi-compartment neuron model according to embodiments of this disclosure.

[0061] The method of this disclosure embodiment is operated in the apparatus for computing the multi-compartment neuron model described above, with reference to... Figure 3 It includes:

[0062] S201, the arithmetic unit uses each corresponding compartment as a virtual neuron to obtain the initial membrane potential of the compartment, and sends the initial membrane potential to the coupling unit.

[0063] S202. The coupling unit calculates the final membrane potential of each chamber based on the initial membrane potential of each chamber in the same multi-compartment neuron model, and sends the final membrane potential to the corresponding computation unit of each chamber.

[0064] In this embodiment of the present disclosure, in each time step, the computing unit (such as a single-neuron computing device) treats each of its corresponding chambers as a neuron (virtual neuron) and calculates the membrane potential (the membrane potential affected by input, discharge, leakage, etc.) that each virtual neuron should have, i.e. the initial membrane potential of the chamber, and sends the initial membrane potential to the coupling unit; the coupling unit performs coupling operation based on the initial membrane potential of each chamber in a multi-chamber neuron model to obtain the final membrane potential of each chamber after coupling, and sends it back to the computing unit corresponding to each chamber, so that each computing unit can obtain the final membrane potential of the corresponding chamber in this time step and complete the operation of the multi-chamber neuron model.

[0065] There are various ways to calculate the initial and final membrane potentials.

[0066] For example, in a certain time frame, the charging current I of the chamber can be calculated based on the input events and related parameters (such as input values, connection weights, etc.) input to the chamber; then, based on the charging current I, the final membrane potential U of the chamber in the previous time frame, and related parameters (such as leakage, discharge threshold, resting potential, etc.), it can be calculated whether the chamber will discharge, and its corresponding initial membrane potential V.

[0067] The final membrane potential U of a compartment can be calculated based on its own and the initial membrane potential V of the adjacent compartments according to the following formula:

[0068]

[0069] Among them, U k V represents the final membrane potential of the k-chamber. k V i V j R represents the initial membrane potentials of commissure k, commissure i, and commissure j, respectively. ik R represents the coupling resistance between chamber i and chamber k. kj Let M represent the coupling resistance between room k and room j, M be the set of the preceding adjacent rooms of room k, and N be the set of the following adjacent rooms of room k.

[0070] In this context, "predecessor" refers to the direction from the ventricle where the axon (output) of the multi-ventricle neuron model is located to the farthest ventricle, while "successor" refers to the direction from one ventricle to the ventricle where the axon (output) of the multi-ventricle neuron model is located; and "adjacent" only includes directly adjacent.

[0071] For example, refer to Figure 1 In the multi-compartment neuron model, the preceding adjacent compartment of compartment C2 is compartment C1, and the subsequent adjacent compartment is compartment C3. (Referencing...) Figure 5 The preceding adjacent rooms of room C7 are rooms C5 and C6, and the following adjacent room is room C8; while room C5 has no preceding adjacent room, and the following adjacent room is room C7.

[0072] In some embodiments, each processing unit corresponds to only one room.

[0073] As one embodiment of this disclosure, each computing unit may correspond to only one room and perform operations only on that room.

[0074] In some embodiments, the method of this disclosure is used to perform operations on multiple multi-compartment neuron models with the same structure; each operation unit corresponds to multiple compartments, and the multiple compartments corresponding to the same operation unit are the compartments in the same relative position in each multi-compartment neuron model.

[0075] As another embodiment of this disclosure, when performing calculations on multiple structurally identical multi-compartment neuron models, one can refer to... Figure 4 In multi-compartment neuron models, compartments in the same relative position are grouped into "a group (or a layer)," for example... Figure 4 Each C1 compartment is grouped into one group, each C2 compartment into another group, each C3 compartment into another group, and each C4 compartment into yet another group. Multiple compartments from different multi-compartment neuron models within the same group correspond to one computational unit, which processes them. Compartments from different groups correspond to different computational units.

[0076] Among them, "multi-compartment neuron models with the same structure" means that the number of compartments, the connection relationship between compartments, and the type of compartments (such as the connection with dendrites and axons) are the same in each multi-compartment neuron model. However, it does not mean that the parameters of compartments in the same relative position in different multi-compartment neuron models are the same, nor does it mean that the connection relationship between each multi-compartment neuron model and other neuron models in the outside world is the same.

[0077] Therefore, the above grouping (layers) does not require that each compartment in the multi-compartment neuron model must be... Figure 4 The "serial" format in the text does not require that the relative positions of the rooms be the same; for example, multiple reference rooms can be used. Figure 5 In the multi-compartment neuron model, multiple C5 compartments can be grouped into one group, while multiple C6 compartments can be grouped into another group.

[0078] In some embodiments, the computational unit uses each corresponding ventricle as a virtual neuron to obtain the initial membrane potential of the ventricle (S201) including:

[0079] S2011. The computational unit uses multidimensional vector operations to obtain the initial membrane potential of each chamber as a virtual neuron.

[0080] In multidimensional vector operations, each room corresponding to the operation unit corresponds to one dimension of the multidimensional vector.

[0081] When a processing unit corresponds to multiple "groups" of compartments, the processing method (but not the specific parameters) for each compartment is the same. Therefore, in order to speed up the processing, each type of data corresponding to each compartment (such as the final membrane potential of the previous row, the input, the initial membrane potential of the current time step, parameters, etc.) can be combined into a multi-dimensional vector. That is, each "dimensional" element in each multi-dimensional vector is a type of data for a compartment, and all elements in each multi-dimensional vector are the same type of data for each compartment in the "group".

[0082] Therefore, the calculation of the initial membrane potential of multiple compartments can be transformed into the calculation of a multidimensional vector, that is, the same operation is performed on multiple elements in the multidimensional vector, and each element in the resulting multidimensional vector is the initial membrane potential of a compartment.

[0083] Moreover, as is known, the operations in each "dimension" of a vector are performed in "parallel". Therefore, according to the embodiments of this disclosure, the same operations can be performed on the multiple chambers of multiple multi-compartment neurons in parallel, such as the membrane potential of these multiple chambers at a certain time beat in parallel, so that multiple multi-compartment neurons can be processed in parallel, and the manner in which they are in the same relative position is also processed in parallel, thereby improving the computation and speed.

[0084] In some embodiments, at least the arithmetic unit corresponding to a compartment having dendrites has an input terminal.

[0085] The computational unit uses each corresponding compartment as a virtual neuron to obtain the initial membrane potential of the compartment (S201), including:

[0086] S2012. When the computational unit corresponding to the atrium with dendrites receives an input signal from its input terminal, the virtual neuron of the corresponding atrium receives the input event and calculates the initial membrane potential of the atrium.

[0087] If the input terminal of the arithmetic unit receives an input signal, it indicates that the corresponding virtual neuron of the atrium with dendrites should have received an input event, and the initial membrane potential of the atrium can be calculated in the manner of receiving the input event (such as a pulse).

[0088] In some embodiments, at least the arithmetic unit corresponding to a compartment having an axon has an output terminal.

[0089] The computational unit uses each corresponding compartment as a virtual neuron to obtain the initial membrane potential of the compartment (S201), including:

[0090] S2013. When it is determined that the virtual neuron corresponding to the atrioventricular junction with an axon is firing, the atrioventricular corresponding computational unit sends an output signal to its output terminal.

[0091] If, based on calculations, it is determined that a virtual neuron in a chamber with an axon should fire (e.g., if the membrane potential exceeds the firing threshold, thus requiring the firing of a pulse), then an output signal is emitted from the output terminal of the computation unit corresponding to that chamber, indicating that the corresponding chamber is firing; of course, the initial membrane potential of the chamber obtained at this time is also the membrane potential after firing.

[0092] When the computing unit performs operations on multiple rooms in the form of a multidimensional vector, the input / output signals it receives / transmits correspond to different elements in the vector through time. For example, in a time period, different sub-time periods correspond to different elements of the vector (that is, different rooms), so the corresponding room can be determined according to the time of the input / output signal.

[0093] Thirdly, referring to Figure 6 This disclosure provides a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements a method for operating any of the multi-compartment neuron models of this disclosure.

[0094] Among them, the processor is a device with data processing capabilities, including but not limited to the central processing unit (CPU); the computer-readable medium is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically such as SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), and flash memory (FLASH); the I / O interface (read-write interface) is connected between the processor and the memory, enabling information exchange between the memory and the processor, including but not limited to the data bus (Bus).

[0095] Those skilled in the art will understand that all or some of the steps, systems, and devices disclosed above, as functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0096] In hardware implementations, the division between functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be executed by several physical components working together.

[0097] Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit (CPU), digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software may be distributed on a computer-readable medium, which may include both computer-readable media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer-readable media includes volatile and non-volatile, removable and non-removable media implemented in any method or technique for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer-readable media include, but are not limited to, random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory (FLASH) or other disk storage; read-only optical disc (CD-ROM), digital versatile disc (DVD) or other optical disc storage; magnetic cartridges, magnetic tapes, disk storage or other magnetic storage; and any other media that can be used to store desired information and can be accessed by a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0098] This disclosure has disclosed exemplary embodiments, and although specific terminology has been used, it is for general illustrative purposes only and should not be construed as limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of this disclosure as set forth by the appended claims.

Claims

1. An apparatus for computing multi-compartment neuron models, wherein each multi-compartment neuron model includes multiple compartments, and the apparatus includes a coupling unit and multiple computing units connected to the coupling unit; wherein, Each of the aforementioned computational units is used to obtain the initial membrane potential of the commissure by using each corresponding commissure as a virtual neuron; multiple commissures of the same multi-commissure neuron model correspond to different computational units; The coupling unit is used to obtain the initial membrane potential obtained by each computing unit, calculate the final membrane potential of each chamber based on the initial membrane potential of each chamber in the same multi-chamber neuron model, and send the final membrane potential to the computing unit corresponding to each chamber. The final membrane potential is calculated according to the following formula: ； Among them, U k V represents the final membrane potential of the k-chamber. k V i V j R represents the initial membrane potentials of commissure k, commissure i, and commissure j, respectively. ik R represents the coupling resistance between chamber i and chamber k. kj Let M represent the coupling resistance between ventricle k and ventricle j, M be the set of preceding adjacent ventricles of ventricle k, and N be the set of succeeding adjacent ventricles of ventricle k. Here, the preceding ventricle refers to the direction from the ventricle where the axon of the multi-ventricle neuron model is located to the farthest ventricle, and the succeeding ventricle refers to the direction from one ventricle to the ventricle where the axon of the multi-ventricle neuron model is located.

2. The apparatus according to claim 1, wherein, At least the computational unit corresponding to a compartment with dendrites has an input terminal.

3. The apparatus according to claim 1, wherein, At least the arithmetic unit corresponding to a compartment with an axon has an output terminal.

4. A method for operating a multi-compartment neuron model, based on the apparatus for operating a multi-compartment neuron model according to any one of claims 1 to 3, the method comprising: The computational unit uses each corresponding compartment as a virtual neuron to calculate the initial membrane potential of the compartment and sends the initial membrane potential to the coupling unit; The coupling unit calculates the final membrane potential of each chamber based on the initial membrane potential of each chamber in the same multi-compartment neuron model, and sends the final membrane potential to the corresponding computation unit of each chamber.

5. The method according to claim 4, wherein, Each of the aforementioned processing units corresponds to only one room.

6. The method according to claim 4, wherein, The method is used to compute multiple multi-compartment neuron models with identical structures; Each of the aforementioned operational units corresponds to multiple compartments, and the multiple compartments corresponding to the same operational unit are the compartments in the same relative position in each multi-compartment neuron model.

7. The method according to claim 6, wherein, The computational unit uses each corresponding compartment as a virtual neuron to obtain the initial membrane potential of the compartment, including: The computational unit uses multidimensional vector operations, treating each chamber as a virtual neuron, to obtain the initial membrane potential of each chamber. In the multidimensional vector operation, each chamber of the computational unit corresponds to one dimension of the multidimensional vector.

8. The method according to claim 6, wherein, At least the computational unit corresponding to a compartment with dendrites has an input terminal; The computational unit calculates the initial membrane potential of each ventricle using each corresponding ventricle as a virtual neuron, including: when the computational unit corresponding to a ventricle with dendrites receives an input signal from its input terminal, it determines that the virtual neuron of the corresponding ventricle has received an input event, and calculates the initial membrane potential of the ventricle.

9. The method according to claim 6, wherein, At least the arithmetic unit corresponding to a compartment with an axon has an output terminal; The computational unit calculates the initial membrane potential of each ventricle as a virtual neuron, including: when it is determined that the virtual neuron corresponding to the ventricle with an axon fires, the computational unit corresponding to the ventricle sends an output signal to its output terminal.

10. A computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the method for operating a multi-compartment neuron model according to any one of claims 4 to 9.

Images