Neuron spike potential classification method, device, storage medium and computer equipment based on parallel superparamagnetic clustering algorithm

Through the application of parallel superparamagnetic clustering algorithm, the problem of low efficiency in the classification of neuron peak potentials in the existing technology is solved, efficient classification of massive data is achieved, and the computing speed is significantly improved.

CN114931388BActive Publication Date: 2024-11-01GUANGDONG MEDICAL UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210447457.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-26
Publication Date
2024-11-01
Estimated Expiration
2042-04-26

AI Technical Summary

Technical Problem

In the prior art, when using superparamagnetic clustering algorithm to classify neuron peak potentials, the calculation efficiency is low and the speed is slow, especially when facing massive data.

Method used

The parallel superparamagnetic clustering algorithm is used to generate Hamiltonian energy that conforms to the Potzmann probability distribution in parallel, and find the closest element in the estimated matrix, filter out the corresponding index set for magneton binding, and conduct statistical probability values. Clustering avoids the probability operation of traversing all magnetons and improves the calculation efficiency.

Benefits of technology

The operation speed of the classification process has been significantly improved. The classification method has improved at least 8.7 times faster on the basis of obtaining similar accuracy, especially when processing massive data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114931388B_ABST
    Figure CN114931388B_ABST
Patent Text Reader

Abstract

The embodiment of the present application discloses a neuron peak potential classification method, device, storage medium and computer equipment based on a parallel superparamagnetic clustering algorithm, which relates to the field of signal classification. The present application adopts a generation system state that can be realized in parallel, so that the calculation of the number of system states required by the algorithm can be performed simultaneously, and the generation of a new state does not need to use probability calculation to traverse all magnetic particles in the system, but is completed by table lookup, and the table lookup process has also been optimized in the algorithm; therefore, the running speed of the classification process can be greatly improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of signal classification, and in particular to a neuron peak potential classification method, device, storage medium and computer equipment based on a parallel superparamagnetic clustering algorithm. Background Art

[0002] The development of neuronal signal recording technology has enabled neuroscientists to record the electrophysiological activity information of individual neurons in the brain, which exists in the form of spikes or action potentials. By analyzing the recorded electrophysiological activity of groups of neurons, the functions of neural circuits or neural networks can be decoded, thus providing the possibility of understanding the functions of the brain. In vivo neuronal signal recording systems generally use extracellular recording. However, extracellular recording measures the collection of electrophysiological activities of several neurons around the electrode. Analyzing these activities requires algorithms to classify which neuron the recorded neuronal signal (spike) is emitted by.

[0003] The superparamagnetic clustering (SPC) algorithm is an unsupervised data clustering algorithm with weak parameter dependence. Due to these excellent characteristics of the algorithm, the algorithm is usually applied to the classification of neuronal spikes. The superparamagnetic clustering algorithm introduces the phenomena and theories of physics into the information field, and abstracts the data clustering process into the thermodynamic aggregation movement of magnons in the inhomogeneous Potts model. In the Potts model, the data to be clustered is regarded as magnons in a magnetic system, and the clustering results are characterized by the system state spontaneously formed by the change of the spin state between magnons at different temperatures: when the temperature changes, the magnons spontaneously synchronize their spins to form different "magnetic clusters" according to the strength of the coupling force between magnons; the "magnetic clusters" spontaneously formed between magnons caused by temperature changes can be regarded as a classification result of the input data by the system. Therefore, the SPC clustering results are closely related to temperature, and different classification results will be presented at different temperatures. This feature enables the SPC algorithm to process some classified data with hierarchical structures: by using different temperatures, the classification structure of data at different levels can be displayed; while traditional clustering algorithms usually only have one classification result. At the same time, compared with traditional clustering algorithms, the SPC algorithm spontaneously generates clustering results based on the characteristics of the data's own structure, so there is no special assumption about the structure of the input data (such as the K-means algorithm assumes that the data structure follows a Gaussian distribution), and it is not highly dependent on various parameters. The implementation of the SPC algorithm is based on quantitative analysis based on the Potts model.

[0004] In the prior art, using the SPC algorithm to classify neuronal spikes is a process of building a Markov chain, in which the generation of the next state depends on the result of the previous state. This sequential execution process greatly reduces the classification efficiency; especially when faced with a large number of neuronal spikes, its shortcomings of slow calculation and inefficiency are further magnified. Summary of the invention

[0005] The embodiments of the present application provide a neuron spike potential classification method, device, storage medium and computer equipment based on a parallel superparamagnetic clustering algorithm, which can solve the problems of low classification efficiency and slow speed in the prior art. The technical solution is as follows:

[0006] In a first aspect, an embodiment of the present application provides a neuron spike potential classification method based on a parallel superparamagnetic clustering algorithm, the method comprising:

[0007] Obtain N neuron spikes to be classified and input them into the magnetic subsystem as N magnetic particles; wherein each neuron spike corresponds to one magnetic particle, and N is an integer greater than 1;

[0008] Calculate the interaction force between each magnon and its nearest K neighbor magnons to obtain a first matrix; wherein K is an integer less than N-1, and the first matrix is ​​composed of elements of N rows and K columns, and each element is an interaction force;

[0009] Arrange the elements in the first matrix in ascending order to obtain a second matrix; wherein the second matrix consists of 1 row and N×K columns of elements;

[0010] Cumulatively summing up each element in the second matrix to obtain an estimation matrix;

[0011] Generate Y Hamiltonian energies in parallel that conform to the Boltzmann probability distribution;

[0012] Searching in parallel the estimated matrix for the element closest to each Hamiltonian energy;

[0013] Determining the two-dimensional index of each element before the closest element;

[0014] Filter out two-dimensional indexes other than the determined two-dimensional index in the index set corresponding to the first matrix;

[0015] Bind the magnetic sub-pairs corresponding to the screened two-dimensional indexes;

[0016] According to the Y binding results, the probability value between any two magnetons is calculated;

[0017] The N neuron spikes are clustered according to the probability values.

[0018] In a second aspect, an embodiment of the present application provides a neuron spike potential classification device based on a parallel superparamagnetic clustering algorithm, the device comprising:

[0019] An acquisition unit, used for acquiring N neuron spikes to be classified and inputting them into the magnetic subsystem as N magnetic particles; wherein each neuron spike corresponds to one magnetic particle, and N is an integer greater than 1;

[0020] A calculation unit, used for calculating the interaction force between each magnon and its nearest K neighbor magnons, to obtain a first matrix; wherein K is an integer less than N-1, and the first matrix is ​​composed of elements of N rows and K columns, and each element is an interaction force;

[0021] A sorting unit, used to arrange the elements in the first matrix in ascending order to obtain a second matrix; wherein the second matrix consists of 1 row and N×K columns of elements;

[0022] A summing unit, used for accumulating and summing each element in the second matrix to obtain an estimated matrix;

[0023] A generating unit, used for generating Y Hamiltonian energies in parallel that conform to the Boltzmann probability distribution;

[0024] A binding unit is used to search the estimated matrix in parallel for the element closest to each Hamiltonian energy; determine the two-dimensional index of each element before the closest element; filter out the two-dimensional indexes other than the determined two-dimensional index in the index set corresponding to the first matrix; and bind the magnetic pairs corresponding to the filtered two-dimensional indexes;

[0025] The clustering unit is used to count the probability value between any two magnetic particles according to the Y binding results; and cluster the N neuron peak potentials according to the probability value.

[0026] In a third aspect, an embodiment of the present application provides a computer storage medium, wherein the computer storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded by a processor and executing the above-mentioned method steps.

[0027] In a fourth aspect, an embodiment of the present application provides a computer device, which may include: a processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above-mentioned method steps.

[0028] The beneficial effects brought about by the technical solutions provided by some embodiments of the present application include at least:

[0029] The classification method uses a parallelizable generation system state, so that operations on the number of system states required by the algorithm can be performed simultaneously. The generation of new states does not require the use of probability operations to traverse all magnetic particles in the system, but is completed through table lookup. The table lookup process has also been algorithmically optimized; therefore, the running speed of the classification process can be greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0031] Figure 1 It is a flow chart of a neuron spike potential classification method based on a parallel superparamagnetic clustering algorithm provided in an embodiment of the present application;

[0032] Figure 2 It is a performance diagram of the classification method provided in the embodiment of the present application;

[0033] Figure 3 It is a structural schematic diagram of a neuron spike potential classification device based on a parallel superparamagnetic clustering algorithm provided by the present application;

[0034] Figure 4 It is a structural schematic diagram of a computer device provided by this application. DETAILED DESCRIPTION

[0035] In order to make the objectives, technical solutions and advantages of the present application more clear, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

[0036] It should be noted that the neuron spike potential classification method based on the parallel superparamagnetic clustering algorithm provided in the present application is generally executed by a computer device, and accordingly, the neuron spike potential classification device based on the parallel superparamagnetic clustering algorithm is generally set in the computer device.

[0037] The computer device may be any computer device with a display screen, including but not limited to a smart phone, a tablet computer, a laptop computer, a desktop computer, etc. When the computer device is software, it may be installed in the computer device listed above. It may be implemented as multiple software or software modules (for example, to provide distributed services), or as a single software or software module, which is not specifically limited here.

[0038] In the related art, the process of clustering neuron spike potentials using the superparamagnetic clustering algorithm includes:

[0039] Assume that the number of neuron spikes to be classified is N, and each neuron spike corresponds to a magneton, that is, there are N magnetons, each of which has D input features and is an input point in the D-dimensional feature space. In each magneton system, there are q possible spin values ​​(spin) to represent its spin state, and the spin states of N magnetons together constitute the system state S. The energy H(S) of the system in state S is calculated using the following formula (1), where H(S) is the Hamiltonian function:

[0040]

[0041] Among them, J ij is the interaction force between magnon i and magnon j, which is generally expressed by formula (2):

[0042]

[0043] Among them, J ij The size is inversely proportional to the distance between the magnons, and a is the average distance between all the magnons. i is the eigenvalue of magneton i, x j is the characteristic value of magnon j. When the spin values ​​of magnon i and magnon j are the same, otherwise That is, only when the spin states of the two magnetons are different, the interaction force J ij It contributes to the calculation of energy H(S). When the spin values ​​are the same, the magnon does not contribute to the calculation of energy.

[0044] The energy distribution of the system at temperature T conforms to the Boltzmann distribution, as shown in formula (3), where Z is a quantization constant and the calculated result is T.

[0045]

[0046] In a magnetic system, the thermodynamic motion of magnons at different temperatures T causes the magnetic system to present different phases. Only the magnetic clusters formed in the superparamagnetic phase (SPC phase) can characterize the clustering results of the system. Therefore, the first step of the SPC algorithm is to estimate the temperature range required for the system to be in the superparamagnetic phase. Usually, we use the Potts model to quantify the physical properties of the magnetic system. In the Potts model, the susceptibility χ is used to estimate the temperature range in which the system is in the SPC phase: when the system gradually heats up from the low-temperature ferromagnetic phase to the superparamagnetic phase, χ will have an obvious peak value; when the system further heats up from the superparamagnetic phase to the paramagnetic phase, the χ value will decrease significantly. Therefore, by calculating the change characteristics of χ with temperature T, it is possible to determine the temperature range in which the system is in the superparamagnetic phase. The change in χ value is caused by the change in the magnetic clusters formed by the spin of magnons with temperature changes. χ can be calculated using formulas (4) and (5):

[0047]

[0048]

[0049] Where m(S) is the magnetization, N max (S) is the maximum number of magnons with the same spin value in the system in state S, and <> represents the mean value.

[0050] At a specific temperature T, the system state S has q N possible composition states, directly calculate <m>It is unrealistic to calculate χ. Generally, the Swendnsen-Wang or Wolf algorithm is used to construct Y sub-states to calculate q. N The total state is sampled, and the Y sub-states generated by sampling are used to estimate the actual <m>And the value of χ. Obviously, the larger the value of Y, the more accurate the estimation result, but the corresponding required calculation amount is larger. Generally, Y≥300 can meet the accuracy requirement. Preferably, Y=300, which can be calculated with fewer samples and reduce the amount of calculation.

[0051] For example: The sampling process implemented by the Wolf algorithm is as follows:

[0052] (1) Determine the number Y of system substates that need to be generated.

[0053] (2) Set an initial spin value (1…q) for the N magnetons in the system to form the initial state S of the system init .

[0054] (3) Randomly select a magneton F from N magnetons and All neighboring magnons of the magnon are traversed to determine whether the neighboring magnons are in a bound state with the magnon F. Among them, only the neighboring magnons with the same spin value participate in the traversal.

[0055] (4) Assign a new random spin value to the magnon F and its bound neighboring magnons. These random spin values ​​constitute the new system state S new .

[0056] (5) Statistics in system state S new The magnetization intensity under (calculated using formula (5)), and the number of states generated by statistics S new Has Y been reached? If not, repeat step (3) until Y S are generated. new Then use the generated Y Snew states to calculate χ and G ij , G ij is the correlation coefficient between magnon i and magnon j, which can generally be expressed by the probability that the two have the same spin state in Y system states.

[0057] By performing the calculations of steps (1) to (5) at different temperatures T, we can obtain a curve of χ changing with temperature T. Then, we can locate the temperature range where the superparamagnetic phase is located based on the change of χ, and thus determine the superparamagnetic phase according to the selected temperature T. select G corresponding to the lower χ value ij Value to perform clustering operation: If G ij >Threshold (the value of Threshold is generally 0.5), then bind magneton i and magneton j. By traversing the G of each magneton ij By deriving the value, we can get the bound magnetic sub-relationship pairs, which will serve as the basis for the final clustering.

[0058] From the above clustering process, we can see that each time a new state S is generated new It is generated based on the previous state and is also the starting state of the next state. new The generation of is a sequential process, which greatly reduces the operating efficiency of the entire system; especially when facing large Y values ​​or massive data, its operating efficiency will appear to be relatively low, and the classification process will consume a lot of time.

[0059] The following will be combined with the attached Figure 1 , the neuron peak potential classification method based on the parallel superparamagnetic clustering algorithm provided in the embodiment of the present application is introduced in detail.

[0060] See also Figure 1 , which is a flow chart of a neuron peak potential classification method based on a parallel superparamagnetic clustering algorithm provided in the present application embodiment. Figure 2 As shown, the method of the embodiment of the present application may include the following steps:

[0061] S101. Obtain N neuron spikes to be classified and input them into a magnetic subsystem as N magnetic subs.

[0062] The computer device obtains N neuron spikes to be classified through a signal acquisition device, where N is an integer greater than 1 and the value of N is generally large. Each neuron spike has D features, where D is an integer greater than or equal to 1, that is, the features of the neuron spike are one-dimensional or multi-dimensional.

[0063] S102. Calculate the interaction force between each magnon and its K nearest neighbor magnons to obtain a first matrix.

[0064] Since the number of magnons is N, and each magnon has N-1 neighboring magnons, if the interaction force between the magnon and the N-1 magnons is calculated, the amount of calculation will be very large. In order to reduce the amount of calculation, for each magnon, the distance between the magnon and the N-1 neighboring magnons is calculated, and then the calculated N-1 distances are sorted from large to small, and the first K corresponding magnons are taken to obtain K neighboring magnons, and then the interaction force between the magnon and the K neighboring magnons is calculated. The method for calculating the interaction force is shown in formula (2), x i represents the characteristic value of the peak potential of the i-th neuron, x j Represents the characteristic value of the jth neuron spike potential.

[0065] The interaction force between each magnon and its nearest K neighbor magnons is calculated to obtain the first matrix. The first matrix has N rows and K columns. Each row corresponds to a magnon, and the K elements in a row are the interaction forces between the magnon and its nearest K neighbor magnons. Each element in the first matrix has a two-dimensional index, which is the serial number of the two magnons.

[0066] S103: Arrange the elements in the first matrix in ascending order to obtain a second matrix.

[0067] Among them, the N×K elements in the first matrix are arranged in order from small to large into a row to obtain the second matrix, the second matrix consists of 1 row and N×K columns of elements, and the two-dimensional index of each element in the second matrix in the first matrix is ​​recorded at the same time.

[0068] S104: Accumulate and sum each element in the second matrix to obtain an estimation matrix.

[0069] Among them, the estimated matrix is ​​composed of 1 row and N×K columns. Let the i-th element in the first matrix be A i , the i-th element in the estimation matrix is ​​B i , then B i =A i +A i-1 +, ..., +A1, that is, the method of cumulative summation is: for the i-th element in the first matrix, the value obtained by summing the i-th element, the i-1-th element, ..., the 1st element replaces the i-th element. For example: for the 3rd element in the second matrix, the value obtained by summing the 3rd element, the 2nd element, and the 1st element replaces the 3rd element.

[0070] S105. Generate Y Hamiltonian energies in parallel that conform to the Boltzmann probability distribution.

[0071] The computer device simultaneously generates Y Hamiltonian energies that conform to the Boltzmann probability distribution. For example, the computer device has N processing cores, and each processing core generates a Hamiltonian energy that conforms to the Boltzmann probability distribution to improve processing efficiency.

[0072] In one or more possible embodiments, Y H(S) can be generated simultaneously by using inverse CDF sampling according to the Boltzmann probability distribution law. i ):

[0073] H(S i )=-T*ln[1-r i ]fori=1...Y.

[0074] Because H(S i ) is the interaction force J of all magnetons in the system ij The next step is how to select Jij so that their sum reaches H(S i ). Select J ij The magnetons have different spin values, and J is not selected. ij The magnetons have the same spin value (bound); through such selection, a new system spin state S can be formed i , so that S i Energy H(S i ) conforms to the Boltzmann probability distribution.

[0075] J ij The specific selection method can refer to the above-mentioned Wolf algorithm: whether the magneton i and the magneton j are bound (i.e., the same spin value) is determined by probability Come to judge, bound J ij is not involved in H(S i ) calculation, the rest of J ij Participate in H(S i ) is calculated, and the final calculated H(S i ) conforms to the Boltzmann probability distribution. Because It is inversely proportional to the distance between the magnetons: that is, the smaller the distance between the magnetons, the greater the probability of binding. Therefore, the J with a high probability of being selected in the Wolf algorithm ij is generated between those magnetons that are farther away. Based on this, this application selects J ij When the J generated between the magnetons with larger distances is directly selected ij : That is, the smaller value J ij (J ij The size of is inversely proportional to the distance between the magnetons). Based on these choices, the total number and probability of binding between magnetons i and j can be directly obtained, so that clustering can be performed using probability.

[0076] Wolf algorithm uses probability To determine whether magneton i and magneton j are bound, the approach of this application is to select a smaller J ij The clustering effect of the two is the same: because the system clustering uses the total binding probability G of magnets i and j ij To judge, because Proportional to Jij and inversely proportional to the magnetic particle distance, the Wolf algorithm is likely to choose the one with a large distance and J ij Small magnetons form H(S i ). However, the G calculated in this application ij The probability is 100%, while the probability calculated by the Wolf algorithm may be 90%, but both are far greater than the threshold value of 0.5, so the final classification results are consistent.

[0077] S106. For each Hamiltonian energy, find the closest element in the estimation matrix.

[0078] Among them, for each of the Y Hamiltonian energies, to find the closest element in the estimated matrix, a total of Y searches need to be performed. The present application can perform Y searches in the estimated matrix in parallel to improve the search efficiency. For any Hamiltonian energy, the closest element among the N×K elements represents the element with the smallest difference.

[0079] S107: Determine the two-dimensional index of each element before the closest element.

[0080] S108: Filter out two-dimensional indexes other than the determined two-dimensional index in the index set corresponding to the first matrix.

[0081] S109, binding the magnetic sub-pairs corresponding to the screened two-dimensional indexes.

[0082] S110. According to the Y binding results, the probability value between any two magnetons is counted.

[0083] S111. Clustering the N neuron spikes according to the probability values.

[0084] Among them, assuming that the closest element is the kth element in the estimated matrix, then the elements before the kth element, that is, the 1st element to the k-1th element, a total of k-1 elements, determine the two-dimensional index of each element in the k-1 elements in the first matrix, the indexes of the N×K elements contained in the first matrix form an index set, and the two-dimensional indexes except the two-dimensional indexes of the k-1 elements in the index set are screened out, and the magnetic sub-pairs corresponding to the screened two-dimensional indexes are bound, and the magnetic sub-pairs corresponding to the two-dimensional indexes of the k-1 elements are not bound. Then, based on the binding results obtained by Y searches, the probability value between any two magnetic sub-pairs is counted.

[0085] For example: For magnon i and magnon j, calculate the probability value G between magnon i and magnon j. ij Assume that Y = 5. According to the fifth binding result, we know that there are 4 bindings between magneton 1 and magneton 2 and 1 non-binding. Then the probability value between magneton 1 and magneton 2 is 4 / 5 = 0.8. Assume that Threshold = 0.5. The calculated probability value G 12 If it is greater than 0.5, then magnon 1 and magnon 2 are classified into the same category.

[0086] The classification method of the embodiment of the present application adopts a system state generation method that can be implemented in parallel, so that the calculation of the number of system states required by the algorithm can be performed simultaneously. The generation of new states does not need to use probability calculations to traverse all magnetic particles in the system, but is completed through table lookup. The table lookup process has also been algorithmically optimized; therefore, the running speed of the classification process can be greatly improved.

[0087] See also Figure 2 The performance diagram of the classification method is shown in Figure 1. Method 1 is the classification method of this application, and method 2 is the classification method of the prior art. On the basis of obtaining similar clustering accuracy, the classification method of this application has a speed improvement of at least 8.7 times. The larger the amount of data that needs to be clustered, the more obvious the effect of accelerating the execution efficiency of the algorithm of this application. This means that even in the face of massive big data applications, the classification method of this application can continue to exert its unique clustering advantages.

[0088] The following is an embodiment of the device of the present application, which can be used to execute the embodiment of the method of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the method of the present application.

[0089] See also Figure 3 , which shows a schematic diagram of the structure of a neuron peak potential classification device based on a parallel superparamagnetic clustering algorithm provided by an exemplary embodiment of the present application, hereinafter referred to as device 3. The device 3 can be implemented as all or part of a computer device through software, hardware, or a combination of both. Device 4 includes: an acquisition unit 301, a calculation unit 302, a sorting unit 303, a summing unit 304, a generation unit 305, a binding unit 306, and a clustering unit 307.

[0090] The acquisition unit 301 is used to acquire N neuron spikes to be classified and input them into the magnetic subsystem as N magnetic particles; wherein each neuron spike corresponds to one magnetic particle, and N is an integer greater than 1;

[0091] The calculation unit 302 is used to calculate the interaction force between each magnon and its nearest K neighbor magnons to obtain a first matrix; wherein K is an integer less than N-1, and the first matrix is ​​composed of elements of N rows and K columns, and each element is an interaction force;

[0092] A sorting unit 303 is used to arrange the elements in the first matrix in ascending order to obtain a second matrix; wherein the second matrix is ​​composed of 1 row and N×K columns of elements;

[0093] A summing unit 304, configured to cumulatively sum each element in the second matrix to obtain an estimated matrix;

[0094] A generating unit 305, used for generating Y Hamiltonian energies in parallel that conform to the Boltzmann probability distribution;

[0095] The binding unit 306 is used to search the estimated matrix in parallel for the element closest to each Hamiltonian energy; determine the two-dimensional index of each element before the closest element; filter out the two-dimensional index other than the determined two-dimensional index in the index set corresponding to the first matrix; and bind the magnetic sub-pairs corresponding to the filtered two-dimensional index;

[0096] The clustering unit 307 is used to count the probability value between any two magnetic particles according to the Y binding results; and cluster the N neuron peak potentials according to the probability value.

[0097] In one or more possible embodiments, the parallel generation of Y Hamiltonian energies conforming to the Boltzmann probability distribution includes:

[0098] In parallel, the inverse transform sampling method is used to generate Y Hamiltonian energies that conform to the Boltzmann probability distribution:

[0099] H(S i )=-T*ln[1-r i ], i = 1...Y; where Y is an integer greater than 1, r i is a random number between 0 and 1, T is the temperature, and ln is the logarithm with base 10.

[0100] In one or more possible embodiments, 5≤K≤10, and Y≥300.

[0101] It should be noted that, when the device 3 provided in the above embodiment executes the neuron spike classification method based on the parallel superparamagnetic clustering algorithm, only the division of the above functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above functions. In addition, the neuron spike classification device based on the parallel superparamagnetic clustering algorithm provided in the above embodiment and the neuron spike classification method based on the parallel superparamagnetic clustering algorithm are of the same concept, and the embodiment and implementation process thereof are detailed in the method embodiment, which will not be repeated here.

[0102] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0103] The present application also provides a computer storage medium, which can store multiple instructions, and the instructions are suitable for being loaded and executed by a processor as described above. Figure 1 The method steps of the embodiment shown in the figure can be found in the specific implementation process. Figure 1 The specific description of the illustrated embodiment will not be repeated here.

[0104] The present application also provides a computer program product, which stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the neuron peak potential classification method based on the parallel superparamagnetic clustering algorithm as described in the above embodiments.

[0105] See also Figure 4 , which is a schematic diagram of the structure of a computer device according to an embodiment of the present application. Figure 4 As shown, the computer device 400 may include: at least one processor 401 , at least one network interface 404 , a user interface 403 , a memory 405 , and at least one communication bus 402 .

[0106] The communication bus 402 is used to realize the connection and communication between these components.

[0107] The user interface 403 may include a display screen (Display) and a camera (Camera), and the optional user interface 403 may also include a standard wired interface and a wireless interface.

[0108] The network interface 404 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0109] Among them, the processor 401 may include one or more processing cores. The processor 401 uses various interfaces and lines to connect various parts within the entire computer device 400, and executes various functions and processes data of the computer device 400 by running or executing instructions, programs, code sets or instruction sets stored in the memory 405, and calling data stored in the memory 405. Optionally, the processor 401 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 401 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 401, and it can be implemented separately through a chip.

[0110] Among them, the memory 405 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 405 includes a non-transitory computer-readable storage medium. The memory 405 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 405 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 405 may optionally be at least one storage device located away from the aforementioned processor 401. As Figure 4 As shown, the memory 405 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an application program.

[0111] exist Figure 4 In the computer device 400 shown in FIG. 1 , the user interface 403 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 401 can be used to call the application program stored in the memory 405 and specifically execute the following steps: Figure 2 The specific process can be referred to Figure 2 As shown, no further description is given here.

[0112] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0113] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.< / m> < / m>

Claims

1. A neuron spike potential classification method based on a parallel superparamagnetic clustering algorithm, characterized in that: include: Obtain N neuron spikes to be classified and input them into the magnetic subsystem as N magnetic particles; wherein each neuron spike corresponds to one magnetic particle, and N is an integer greater than 1; Calculate the interaction force between each magnon and its nearest K neighbor magnons to obtain a first matrix; wherein K is an integer less than N-1, and the first matrix is ​​composed of elements of N rows and K columns, and each element is an interaction force; Arrange the elements in the first matrix in ascending order to obtain a second matrix; wherein the second matrix consists of 1 row and N×K columns of elements; Cumulatively summing up each element in the second matrix to obtain an estimation matrix; In parallel, the inverse transform sampling method is used to generate Y Hamiltonian energies H(S) that conform to the Boltzmann probability distribution. i ); H(S i ) is the interaction force J of all magnetons in the system ij The sum of the two, whether the magneton i and the magneton j are bound is determined by probability, and the bound J ij is not involved in H(S i ) calculation, the rest of J ij Participate in H(S i ) calculation; the final calculated H(S i ) conforms to the Boltzmann probability distribution; it is inversely proportional to the distance between the magnetons: that is, the smaller the distance between the magnetons, the greater the probability of binding; when choosing J ij When the J generated between the magnetons with larger distances is directly selected ij :The smaller the value of J ij ; Based on these choices, the total number and probability of binding between magneton i and magneton j can be directly obtained, so that clustering can be performed using probability; Searching in parallel the estimated matrix for the element closest to each Hamiltonian energy; Determining the two-dimensional index of each element before the closest element; Filter out two-dimensional indexes other than the determined two-dimensional index in the index set corresponding to the first matrix; Bind the magnetic sub-pairs corresponding to the screened two-dimensional indexes; According to the Y binding results, the probability value between any two magnetons is calculated; The N neuron spikes are clustered according to the probability values.

2. The method according to claim 1, characterized in that 5≤K≤10,Y≥300.

3. A neuron spike potential classification device based on a parallel superparamagnetic clustering algorithm, characterized in that: include: An acquisition unit, used for acquiring N neuron spikes to be classified and inputting them into the magnetic subsystem as N magnetic particles; wherein each neuron spike corresponds to one magnetic particle, and N is an integer greater than 1; A calculation unit, used for calculating the interaction force between each magnon and its nearest K neighbor magnons, to obtain a first matrix; wherein K is an integer less than N-1, and the first matrix is ​​composed of elements of N rows and K columns, and each element is an interaction force; A sorting unit, used to arrange the elements in the first matrix in ascending order to obtain a second matrix; wherein the second matrix consists of 1 row and N×K columns of elements; A summing unit, used for accumulating and summing each element in the second matrix to obtain an estimated matrix; A generating unit is used to generate Y Hamiltonian energies H(S) in parallel by using the inverse transform sampling method in accordance with the Boltzmann probability distribution. i ); H(S i ) is the interaction force J of all magnetons in the system ij The sum of the two, whether the magneton i and the magneton j are bound is determined by probability, and the bound J ij is not involved in H(S i ) calculation, the rest of J ij Participate in H(S i ) calculation; the final calculated H(S i ) conforms to the Boltzmann probability distribution; it is inversely proportional to the distance between the magnetons: that is, the smaller the distance between the magnetons, the greater the probability of binding; when choosing J ij When the J generated between the magnetons with larger distances is directly selected ij :The smaller the value of J ij ; Based on these choices, the total number and probability of binding between magneton i and magneton j can be directly obtained, so that clustering can be performed using probability; A binding unit is used to search the estimated matrix in parallel for the element closest to each Hamiltonian energy; determine the two-dimensional index of each element before the closest element; filter out the two-dimensional indexes other than the determined two-dimensional index in the index set corresponding to the first matrix; and bind the magnetic pairs corresponding to the filtered two-dimensional indexes; The clustering unit is used to count the probability value between any two magnetic particles according to the Y binding results; and cluster the N neuron peak potentials according to the probability value.

4. A computer storage medium, characterized in that: The computer storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method steps as claimed in any one of claims 1 to 2.

5. A computer device, characterized in that: include: A processor and a memory; wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the method steps as claimed in any one of claims 1 to 2.