CN116205276A - Apparatus and method for multi-compartment neuron model operation and computer readable medium - Google Patents
Apparatus and method for multi-compartment neuron model operation and computer readable medium Download PDFInfo
- Publication number
- CN116205276A CN116205276A CN202111441571.5A CN202111441571A CN116205276A CN 116205276 A CN116205276 A CN 116205276A CN 202111441571 A CN202111441571 A CN 202111441571A CN 116205276 A CN116205276 A CN 116205276A
- Authority
- CN
- China
- Prior art keywords
- compartment
- membrane potential
- neuron
- chamber
- operation unit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 210000002569 neuron Anatomy 0.000 title claims abstract description 103
- 238000000034 method Methods 0.000 title claims abstract description 31
- 239000012528 membrane Substances 0.000 claims abstract description 103
- 230000008878 coupling Effects 0.000 claims abstract description 32
- 238000010168 coupling process Methods 0.000 claims abstract description 32
- 238000005859 coupling reaction Methods 0.000 claims abstract description 32
- 210000002837 heart atrium Anatomy 0.000 claims abstract description 4
- 210000003050 axon Anatomy 0.000 claims description 20
- 210000001787 dendrite Anatomy 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 210000005056 cell body Anatomy 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000000284 resting effect Effects 0.000 description 2
- 210000000225 synapse Anatomy 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000010304 firing Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000012421 spiking Methods 0.000 description 1
- 230000007723 transport mechanism Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Neurology (AREA)
- Feedback Control In General (AREA)
Abstract
The present disclosure provides an apparatus for multi-compartment neuron model operation, comprising a coupling unit and a plurality of operation units connected with the coupling unit; each operation unit is used for obtaining the initial membrane potential of the atrioventricular by using each corresponding atrioventricular as a virtual neuron; a plurality of atria of the same multi-atria neuron model correspond to different operation units; the coupling unit is used for acquiring the initial membrane potential obtained by each operation unit, obtaining the final membrane potential of each chamber according to the initial membrane potential operation of each chamber of the same multi-chamber neuron model, and sending the final membrane potential to the operation unit corresponding to each chamber. The disclosure also provides a method, an electronic device, and a computer-readable medium for multi-compartment neuron model operation.
Description
Technical Field
The present disclosure relates to the field of neural network technology, and in particular, to an apparatus and method for multi-compartment neuron model operation, and a computer readable medium.
Background
Neural networks such as impulse neural networks (SNN, spiking Neural Networks) have been widely used in the fields of image processing, speech recognition, and the like. The neural network may include a plurality of interconnected neuron models, and the neuron models may receive pulses from other neuron models and pulse the other neuron models when their own membrane potential exceeds a firing threshold.
In order to make the neuron model more accurate to simulate biological neurons, a "multi-chamber neuron model" may be used, that is, referring to fig. 1, a neuron model is divided into a plurality of "chambers", so as to obtain a multi-chamber neuron model formed by a plurality of chambers, where each chamber has a respective membrane potential, and the membrane potentials of different chambers may affect each other.
Therefore, efficient and accurate operation of the multi-compartment neuron model is desirable.
Disclosure of Invention
The present disclosure provides an apparatus and method, computer readable medium for multi-compartment neuron model operation.
In a first aspect, embodiments of the present disclosure provide an apparatus for computing a multi-compartment neuron model, each of the multi-compartment neuron models including a plurality of compartments, the apparatus including a coupling unit and a plurality of computing units connected to the coupling unit; wherein,,
each operation unit is used for obtaining the initial membrane potential of the atrioventricular by using each corresponding atrioventricular as a virtual neuron; a plurality of atria of the same multi-atria neuron model correspond to different operation units;
the coupling unit is used for acquiring the initial membrane potential obtained by each operation unit, obtaining the final membrane potential of each chamber according to the initial membrane potential operation of each chamber of the same multi-chamber neuron model, and sending the final membrane potential to the operation unit corresponding to each chamber.
In some embodiments, at least the arithmetic unit corresponding to the compartment with dendrites has an input.
In some embodiments, at least the arithmetic unit corresponding to the compartment with axons has an output.
In a second aspect, embodiments of the present disclosure provide a method of multi-compartment neuron model operation, performed based on an apparatus of any one of the embodiments of the present disclosure, the method comprising:
the operation unit calculates an initial membrane potential of each chamber by taking each chamber as a virtual neuron, and sends the initial membrane potential to the coupling unit;
the coupling unit obtains the final membrane potential of each compartment according to the initial membrane potential operation of each compartment of the same multi-compartment neuron model, and sends the final membrane potential to the operation unit corresponding to each compartment.
In some embodiments, each of the computing units corresponds to only one compartment.
In some embodiments, the methods of embodiments of the present disclosure are used to operate on multiple structurally identical multi-compartment neuron models;
each operation unit corresponds to a plurality of atrioventricular, and the plurality of atrioventricular corresponding to the same operation unit are the atrioventricular at the same relative position in each multi-atrioventricular neuron model.
In some embodiments, the calculating unit calculates an initial membrane potential of each compartment as a virtual neuron, including:
the operation unit obtains the initial membrane potential of each corresponding chamber by using each corresponding chamber as a virtual neuron in a multidimensional vector operation mode; in the multidimensional vector operation, each room corresponding to the operation unit corresponds to one dimension of the multidimensional vector.
In some embodiments, at least the arithmetic unit corresponding to the compartment with dendrites has an input;
the operation unit calculates the initial membrane potential of the room by taking each corresponding room as a virtual neuron, and the initial membrane potential comprises the following steps: when the operation unit corresponding to the compartment with dendrite receives an input signal from its input end, it is determined that the virtual neuron of the corresponding compartment receives an input event to operate to obtain an initial membrane potential of the compartment.
In some embodiments, at least the arithmetic unit corresponding to the compartment with axons has an output;
the operation unit calculates the initial membrane potential of the room by taking each corresponding room as a virtual neuron, and the initial membrane potential comprises the following steps: when determining that the virtual neuron corresponding to the atrioventricular system with the axon is released, the operation unit corresponding to the atrioventricular system sends out an output signal to the output end of the operation unit.
In a third aspect, the disclosed embodiments provide a computer readable medium having stored thereon a computer program which, when executed by a processor, implements a method of any of the multi-compartment neuron model operations of the disclosed embodiments.
In the embodiment of the disclosure, in a plurality of operation units, each chamber of a multi-chamber neuron model is used as a virtual neuron to calculate and obtain initial membrane potential, and then in a coupling unit, the initial membrane potential of each chamber is coupled to obtain final membrane potential of each chamber; therefore, the embodiment of the disclosure can efficiently and accurately realize the operation of the multi-chamber neuron model through simple devices and calculation.
Drawings
The accompanying drawings are included to provide a further understanding of the disclosure, and are incorporated in and constitute a part of this specification, illustrate the disclosure and together with the detailed embodiment, not to limit the disclosure. The above and other features and advantages will become more readily apparent to those skilled in the art by describing the detailed embodiments with reference to the accompanying drawings in which:
FIG. 1 is a schematic diagram of a multi-compartment neuron model;
FIG. 2 is a block diagram of an apparatus for multi-compartment neuron model operation according to an embodiment of the present disclosure;
FIG. 3 is a flow chart of a method of multi-compartment neuron model operation provided by an embodiment of the present disclosure;
FIG. 4 is a schematic diagram of grouping compartments of a plurality of multi-compartment neuron models in a method of multi-compartment neuron model operation provided by an embodiment of the present disclosure;
FIG. 5 is a schematic diagram of another multi-compartment neuron model;
fig. 6 is a block diagram of a computer readable medium according to an embodiment of the present disclosure.
Detailed Description
For a better understanding of the technical solutions of the present disclosure, the following describes in detail the apparatus and method for multi-compartment neuron model operation, and the computer readable medium provided in the present disclosure with reference to the accompanying drawings.
The present disclosure will be described more fully hereinafter with reference to the accompanying drawings, but the embodiments shown may be embodied in different forms and should not be construed as limited to the embodiments set forth in the disclosure. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
The present disclosure may be described with reference to plan and/or cross-sectional views with the aid of idealized schematic diagrams of the present disclosure. Accordingly, the example illustrations may be modified in accordance with manufacturing techniques and/or tolerances.
Embodiments of the disclosure and features of embodiments may be combined with each other without conflict.
The terminology used in the present disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The term "and/or" as used in this disclosure includes any and all combinations of one or more of the associated listed items. As used in this disclosure, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. The terms "comprises," "comprising," "including," "includes," "including," "having," "including," "made of … …" and/or "comprising," when used in this disclosure, specify the presence of stated features, integers, steps, operations, elements, and/or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and/or groups thereof.
Unless otherwise defined, all terms (including technical and scientific terms) used in this disclosure have the same meaning as commonly understood by one of ordinary skill in the art. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The present disclosure is not limited to the embodiments shown in the drawings, but includes modifications of the configuration formed based on the manufacturing process. Thus, the regions illustrated in the figures have schematic properties and the shapes of the regions illustrated in the figures illustrate the particular shapes of the regions of the elements, but are not intended to be limiting.
In a first aspect, embodiments of the present disclosure provide an apparatus for multi-compartment neuron model operation.
The multi-compartment neuron model according to the embodiment of the present disclosure is used for operating on multi-compartment neuron models, and referring to fig. 1 and 5, each multi-compartment neuron model includes a plurality of compartments.
Referring to fig. 2, an apparatus for multi-compartment neuron model operation according to an embodiment of the present disclosure includes a coupling unit and a plurality of operation units connected to the coupling unit.
Each operation unit is used for obtaining the initial membrane potential of the atrioventricular by using each corresponding atrioventricular as a virtual neuron; multiple atria of the same multi-atrioventricular neuron model correspond to different arithmetic units.
The coupling unit is used for acquiring the initial membrane potential obtained by each operation unit, obtaining the final membrane potential of each chamber according to the initial membrane potential operation of each chamber of the same multi-chamber neuron model, and sending the final membrane potential to the operation unit corresponding to each chamber.
Embodiments of the present disclosure include a plurality of arithmetic units, each of which is a relatively independent physical device having an arithmetic capability, such as a device (e.g., a single neuron arithmetic device) for processing a conventional neuron model.
Moreover, each computing unit corresponds to one or more compartments, and may process each compartment corresponding thereto as a "neuron (virtual neuron", i.e., calculate the membrane potential of the virtual neuron, i.e., calculate the initial membrane potential of the compartment, based on the input of the virtual neuron (e.g., whether a pulse is issued, whether a pulse is received, etc.) during the present time, and the parameters of the virtual neuron (e.g., leakage, issuance threshold, resting potential, etc.). In other words, the "initial membrane potential" refers to the membrane potential that the atrioventricular should have in this time beat if each atrioventricular is an independent neuron (the potentials between the atrioventricular do not affect each other).
Where "time beat" is the minimum time period for which the multi-compartment neuron model operates, i.e., the membrane potential of each compartment of the multi-compartment neuron model is updated once per time beat.
Obviously, since the plurality of compartments of the same multi-compartment neuron model correspond to different operation units, in one time beat, the initial membrane potential of each compartment of each multi-compartment neuron model can be obtained by "synchronizing" the operation of the corresponding plurality of operation units.
In a multi-chamber neuron model, each chamber is not an independent neuron because the potentials of the chambers affect each other in one beat of time, for example, see FIG. 1, where the membrane potential of the C1 chamber affects the membrane potential of the C2 chamber, and the membrane potential of the C2 chamber affects the membrane potential of the C1 chamber; moreover, this effect is not performed in "pulses" but in "analog" fashion; this effect is known as "coupling" of the atrioventricular membrane potential.
The coupling unit is further provided in the embodiment of the present disclosure, and the coupling unit is connected to each operation unit, so that in each time beat, the coupling unit may obtain the initial membrane potential of each chamber of a multi-chamber neuron model obtained by the operation of each operation unit, and calculate, according to the initial membrane potential, the final membrane potential that each chamber should have after coupling, that is, calculate, obtain the final membrane potential that each chamber of the multi-chamber neuron model in the time beat should have, thereby completing the operation of the multi-chamber neuron model.
In the embodiment of the disclosure, in a plurality of operation units, each chamber of a multi-chamber neuron model is used as a virtual neuron to calculate and obtain initial membrane potential, and then in a coupling unit, the initial membrane potential of each chamber is coupled to obtain final membrane potential of each chamber; therefore, the embodiment of the disclosure can efficiently and accurately realize the operation of the multi-chamber neuron model through simple devices and calculation.
In some embodiments, at least the arithmetic unit corresponding to the compartment with dendrites has an input.
In some embodiments, at least the arithmetic unit corresponding to the compartment with axons has an output.
Referring to fig. 1, each neuron model has one or more dendrites (Dendrite) for input thereto, and one Axon (Axon) for output, each Axon being connectable to dendrites of one or more other neurons by synapses (synapses).
In a multi-compartment neuron model, since the neuron model is divided into different compartments, dendrites and axons will actually be located in different compartments, so that the corresponding inputs and outputs also correspond to different compartments, or input events (e.g., pulses) are actually input into a compartment, and outputs (e.g., pulses) are actually generated from a compartment.
Because the compartments are all corresponding to the operation units, if one operation unit corresponds to the compartment with dendrite, the operation unit needs to have an input end to acquire an input signal from the input end, which indicates that the corresponding compartment has input; if one of the computing units has an axon, the computing unit needs to have an output end, so as to send out an output signal from the output end when the corresponding atrioventricular system is about to be sent out.
Of course, there may be some compartments corresponding to only the cell body (Soma), so that there is neither dendrite nor axon, and there is no direct input or output of such compartments, and the membrane potential is determined only by leakage, coupling, etc.
In this case, the input end and the output end may be set for each operation unit according to the condition of the corresponding room, but referring to fig. 2, for the sake of unification of the structure and improving the versatility, all operation units may have the same structure and have the input end and the output end, but if the room corresponding to the operation unit does not include dendrite/axon, the input end/output end of the corresponding operation unit may be "invalid", that is, not connected to other structures.
Of course, when the device for computing the multi-compartment neuron model processes a plurality of multi-compartment neuron models having a connection relationship, the input end and the output end of each computing unit should be connected with each other according to the connection relationship between compartments of different multi-compartment neuron models.
Alternatively, if the device for multi-compartment neuron model operation processes only one multi-compartment neuron model, the input and output terminals in the device for multi-compartment neuron model operation may be connected to each other.
In a second aspect, embodiments of the present disclosure provide a method of multi-compartment neuron model operation, based on an apparatus of any one of the embodiments of the present disclosure.
The method of an embodiment of the present disclosure operates in the apparatus of the above multi-compartment neuron model operation, referring to fig. 3, which includes:
s201, the operation unit calculates to obtain initial membrane potential of the atrioventricular by taking each corresponding atrioventricular as a virtual neuron, and sends the initial membrane potential to the coupling unit.
And S202, the coupling unit calculates the final membrane potential of each compartment according to the initial membrane potential of each compartment of the same multi-compartment neuron model, and sends the final membrane potential to the corresponding operation unit of each compartment.
In the embodiment of the disclosure, in each time beat, the operation unit (such as a single neuron operation device) takes each corresponding chamber as a neuron (virtual neuron), calculates the due membrane potential of each virtual neuron (the membrane potential after being affected by input, release, leakage and the like), namely the initial membrane potential of the chamber, and sends the initial membrane potential to the coupling unit; the coupling unit performs coupling operation according to 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 the final membrane potential back to the corresponding operation unit of each chamber, so that each operation unit can obtain the final membrane potential of the corresponding chamber in the time beat, and the operation of the multi-chamber neuron model is completed.
The specific manner in which the initial and final membrane potentials are calculated is varied.
For example, in a certain time beat, the charging current I of the chamber can be calculated according to the input event and related parameters (such as input value, connection weight, etc.) input to the chamber; and calculating whether the chamber will release or not according to the charging current I, the final membrane potential U of the chamber which is shot at the previous time and related parameters (such as leakage, release threshold, resting potential and the like), and the corresponding initial membrane potential V.
While the final membrane potential U of a compartment can be calculated from its own and the initial membrane potential V of the adjacent compartment according to the following formula:
wherein U is k Represents the final membrane potential of the k chamber, V k 、V i 、V j Respectively represent initial membrane potential of k chamber, i chamber and j chamber, R ik Representing the coupling resistance between i and k compartments, R kj Representing the coupling resistance between k and j compartments, M being the set of successive adjacent compartments of k compartments, N representing the set of successive adjacent compartments of k compartments.
Wherein, the above "successor" refers to the direction pointing from the compartment where the axon (output) of the multi-compartment neuron model is located to the compartment furthest away, and the above "successor" refers to the direction pointing from one compartment to the compartment where the axon (output) of the multi-compartment neuron model is located; whereas the above "adjacent" includes only the immediate adjacency.
For example, referring to the multi-compartment neuron model in FIG. 1, the next adjacent compartment of its C2 compartment is the C1 compartment and the next adjacent compartment is the C3 compartment. And referring to fig. 5, the successive adjacent chambers of the C7 chamber are a C5 chamber and a C6 chamber, and the successive adjacent chamber is a C8 chamber; and for the C5 chamber, there is no preceding adjacent chamber, the following adjacent chamber being the C7 chamber.
In some embodiments, each arithmetic unit corresponds to only one compartment.
As a way of the embodiment of the present disclosure, each operation unit may correspond to only one chamber and perform only the operation of the chamber.
In some embodiments, the methods of embodiments of the present disclosure are used to operate on multiple structurally identical multi-compartment neuron models; each operation unit corresponds to a plurality of atrioventricular, and the plurality of atrioventricular corresponding to the same operation unit are the atrioventricular at the same relative position in each multi-atrioventricular neuron model.
As another way of an embodiment of the present disclosure, when a plurality of multi-chamber neuron models with the same structure are to be operated, chambers at the same relative positions in each multi-chamber neuron model may be divided into "one group (or one layer)", for example, each C1 chamber is one group, each C2 chamber is one group, each C3 chamber is one group, and each C4 chamber is one group in fig. 4, referring to fig. 4; the method comprises the steps that a plurality of chambers from different multi-chamber neuron models in the same group correspond to one operation unit, and the operation unit is used for processing; and different sets of rooms correspond to different operation units.
The "multi-chamber neuron model with the same structure" refers to that the number of chambers, the connection relationship between chambers, the types of chambers (such as connection with dendrites and axons) and the like of each multi-chamber neuron model are the same, but the parameters of chambers at the same relative position in different multi-chamber neuron models are not the same, and the connection relationship between each multi-chamber neuron model and other external neuron models is not the same.
Thus, the above grouping (layer) does not require that each compartment of the multi-compartment neuron model be in the "serial" form of FIG. 4, but rather that the relative positions of compartments be the same, e.g., in the multi-compartment neuron model of multiple reference to FIG. 5, multiple C5 compartments may be grouped together and multiple C6 compartments grouped together.
In some embodiments, the computing unit computes an initial membrane potential of the atrioventricular for each of its corresponding atrioventricular cells as one virtual neuron (S201) comprising:
s2011, the operation unit obtains the initial membrane potential of each corresponding chamber by using each corresponding chamber as a virtual neuron operation in a multidimensional vector operation mode.
In the multidimensional vector operation, each room corresponding to the operation unit corresponds to one dimension of the multidimensional vector.
When one operation unit corresponds to a plurality of 'same group' of rooms, the operation mode (but not specific parameters) of each room is the same, so in order to increase the processing speed, each data (such as the final membrane potential of the previous row, the input, the initial membrane potential of the present time beat, the parameters and the like) corresponding to each room can be formed into a multi-dimensional vector, that is, each 'dimensional' element in each multi-dimensional vector is one data of one room, and all elements in each multi-dimensional vector are the same kind of data of each room in the 'same group'.
Thus, the operation of the initial membrane potential of a plurality of compartments can be converted into the operation of a multidimensional vector, namely, the same operation is carried out on a plurality of elements in the multidimensional vector, and each element in the multidimensional vector obtained by the operation is the initial membrane potential of one compartment.
Moreover, it is known that each "dimensional" operation in a vector is performed "in parallel", whereby, according to embodiments of the present disclosure, multiple ones of multiple multi-compartment neurons can be subjected to the same operation in parallel, such as parallel operation of the membrane potential of the multiple compartments taken at a certain time, so that multiple multi-compartment neurons can be processed in parallel, and wherein the same relative positions are processed in parallel, increasing operation and speed.
In some embodiments, at least the arithmetic unit corresponding to the compartment with dendrites has an input.
The operation unit calculates an initial membrane potential of an atrioventricular by using each corresponding atrioventricular as a virtual neuron (S201) includes:
s2012, when the operation unit corresponding to the compartment with dendrite receives the input signal from the input end thereof, it determines that the virtual neuron of the corresponding compartment receives the input event to calculate to obtain the initial membrane potential of the compartment.
If the input end of the operation unit receives an input signal, the virtual neuron of the corresponding atrioventricular system with dendrite is indicated to receive an input event, so that the initial membrane potential of the atrioventricular system can be calculated according to the mode of receiving the input event (such as pulse).
In some embodiments, at least the arithmetic unit corresponding to the compartment with axons has an output.
The operation unit calculates an initial membrane potential of an atrioventricular by using each corresponding atrioventricular as a virtual neuron (S201) includes:
s2013, when the virtual neuron corresponding to the atrioventricular system with the axon is determined to be released, the operation unit corresponding to the atrioventricular system sends out an output signal to the output end of the operation unit.
If it is determined that a virtual neuron of a certain chamber with an axon should be issued (for example, a membrane potential exceeds an issuing threshold value, so that a pulse should be issued) according to the calculation, an output signal is issued from an output end of a computing unit corresponding to the chamber, so that the corresponding chamber is indicated to be issued; of course, the initial membrane potential of the atrioventricular system calculated at this time is also the membrane potential after issuance.
When the operation unit performs operation on the plurality of compartments in the form of a multidimensional vector, the passing time of the input signal/output signal received/sent by the operation unit corresponds to different elements in the vector, for example, in a time period, different sub-time periods correspond to different elements of the vector (namely correspond to different compartments), so that the corresponding compartments can be determined according to the time of the input signal/output signal.
In a third aspect, referring to fig. 6, the disclosed embodiments provide a computer readable medium having a computer program stored thereon, which when executed by a processor, implements a method of any one of the multi-compartment neuron model operations of the disclosed embodiments.
Wherein the processor is a device having data processing capabilities including, but not limited to, a Central Processing Unit (CPU) or the like; and computer readable media are devices with data storage capabilities including, but 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); the I/O interface (read/write interface) is connected between the processor and the memory, and can implement information interaction between the memory and the processor, which includes, but is not limited to, a data Bus (Bus), etc.
Those of ordinary skill in the art will appreciate that all or some of the steps, systems, functional modules/units in the apparatus disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof.
In a hardware implementation, the division between the functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be performed cooperatively by several physical components.
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. Such software may be distributed on computer readable media, which may include computer readable media (or non-transitory media) and communication media (or transitory media). The term computer readable media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as known to those skilled in the art. Computer-readable media includes, but is not limited to, random access memory (RAM, more particularly SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), FLASH memory (FLASH), or other magnetic disk storage; a compact disk read-only (CD-ROM), digital Versatile Disk (DVD) or other optical disk storage; magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage; any other medium that can be used to store the desired information and that can be accessed by a computer. Furthermore, as is well known to those of ordinary skill in the art, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media.
The present disclosure has disclosed example embodiments, and although specific terms are employed, they are used and should be interpreted in a generic and descriptive sense only and not for purpose of limitation. In some instances, it will be apparent to one skilled in the art that features, characteristics, and/or elements described in connection with a particular embodiment may be used alone or in combination with other embodiments unless explicitly stated otherwise. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the scope of the disclosure as set forth in the appended claims.
Claims (10)
1. An apparatus for multi-compartment neuron model operation, each of the multi-compartment neuron models comprising a plurality of compartments, the apparatus comprising a coupling unit and a plurality of operation units connected to the coupling unit; wherein,,
each operation unit is used for obtaining the initial membrane potential of the atrioventricular by using each corresponding atrioventricular as a virtual neuron; a plurality of atria of the same multi-atria neuron model correspond to different operation units;
the coupling unit is used for acquiring the initial membrane potential obtained by each operation unit, obtaining the final membrane potential of each chamber according to the initial membrane potential operation of each chamber of the same multi-chamber neuron model, and sending the final membrane potential to the operation unit corresponding to each chamber.
2. The apparatus of claim 1, wherein,
at least the arithmetic unit corresponding to the compartment with dendrites has an input.
3. The apparatus of claim 1, wherein,
at least the arithmetic unit corresponding to the compartment with the axon has an output.
4. A method of multi-compartment neuron model operation based on the apparatus of multi-compartment neuron model operation according to any one of claims 1 to 3, the method comprising:
the operation unit calculates an initial membrane potential of each chamber by taking each chamber as a virtual neuron, and sends the initial membrane potential to the coupling unit;
the coupling unit obtains the final membrane potential of each compartment according to the initial membrane potential operation of each compartment of the same multi-compartment neuron model, and sends the final membrane potential to the operation unit corresponding to each compartment.
5. The method of claim 4, wherein,
each operation unit corresponds to only one room.
6. The method of claim 4, wherein,
the method is used for calculating a plurality of multi-chamber neuron models with the same structure;
each operation unit corresponds to a plurality of atrioventricular, and the plurality of atrioventricular corresponding to the same operation unit are the atrioventricular at the same relative position in each multi-atrioventricular neuron model.
7. The method of claim 6, wherein the computing unit computes an initial membrane potential of the compartments with each of its corresponding compartments as one virtual neuron, comprising:
the operation unit obtains the initial membrane potential of each corresponding chamber by using each corresponding chamber as a virtual neuron in a multidimensional vector operation mode; in the multidimensional vector operation, each room corresponding to the operation unit corresponds to one dimension of the multidimensional vector.
8. The method of claim 6, wherein,
at least the arithmetic unit corresponding to the compartment with dendrites has an input;
the operation unit calculates the initial membrane potential of the room by taking each corresponding room as a virtual neuron, and the initial membrane potential comprises the following steps: when the operation unit corresponding to the compartment with dendrite receives an input signal from its input end, it is determined that the virtual neuron of the corresponding compartment receives an input event to operate to obtain an initial membrane potential of the compartment.
9. The method of claim 6, wherein,
at least the arithmetic unit corresponding to the compartment with the axon has an output;
the operation unit calculates the initial membrane potential of the room by taking each corresponding room as a virtual neuron, and the initial membrane potential comprises the following steps: when determining that the virtual neuron corresponding to the atrioventricular system with the axon is released, the operation unit corresponding to the atrioventricular system sends out an output signal to the output end of the operation unit.
10. A computer readable medium having stored thereon a computer program which when executed by a processor implements a method of multi-compartment neuron model operation according to any one of claims 4 to 9.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111441571.5A CN116205276A (en) | 2021-11-30 | 2021-11-30 | Apparatus and method for multi-compartment neuron model operation and computer readable medium |
| PCT/CN2022/131288 WO2023098449A1 (en) | 2021-11-30 | 2022-11-11 | Multi-compartment neuron model operation apparatus and method, and computer-readable medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111441571.5A CN116205276A (en) | 2021-11-30 | 2021-11-30 | Apparatus and method for multi-compartment neuron model operation and computer readable medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116205276A true CN116205276A (en) | 2023-06-02 |
Family
ID=86517800
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111441571.5A Pending CN116205276A (en) | 2021-11-30 | 2021-11-30 | Apparatus and method for multi-compartment neuron model operation and computer readable medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN116205276A (en) |
| WO (1) | WO2023098449A1 (en) |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8515885B2 (en) * | 2010-10-29 | 2013-08-20 | International Business Machines Corporation | Neuromorphic and synaptronic spiking neural network with synaptic weights learned using simulation |
| CN104620236B (en) * | 2012-03-15 | 2019-02-15 | 高通股份有限公司 | The device and method based on label for neural network |
| US11126913B2 (en) * | 2015-07-23 | 2021-09-21 | Applied Brain Research Inc | Methods and systems for implementing deep spiking neural networks |
| AU2016381015B2 (en) * | 2015-12-30 | 2019-04-18 | Boston Scientific Neuromodulation Corporation | System for composing neurostimulation patterns using a neuronal network model |
-
2021
- 2021-11-30 CN CN202111441571.5A patent/CN116205276A/en active Pending
-
2022
- 2022-11-11 WO PCT/CN2022/131288 patent/WO2023098449A1/en unknown
Also Published As
| Publication number | Publication date |
|---|---|
| WO2023098449A1 (en) | 2023-06-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108336427B (en) | Battery management method and device | |
| US4979126A (en) | Neural network with non-linear transformations | |
| Amari | Dualistic geometry of the manifold of higher-order neurons | |
| US5544336A (en) | Parallel data processing system which efficiently performs matrix and neurocomputer operations, in a negligible data transmission time | |
| CN112734012B (en) | Impulse neural network training method, data processing method, electronic equipment and medium | |
| CN111723901A (en) | Training method and device of neural network model | |
| CN110930996A (en) | Model training method, voice recognition method, device, storage medium and equipment | |
| CN117634564B (en) | Pulse delay measurement method and system based on programmable nerve mimicry core | |
| CN111260056A (en) | Network model distillation method and device | |
| CN116205276A (en) | Apparatus and method for multi-compartment neuron model operation and computer readable medium | |
| CN109697511A (en) | Data reasoning method, apparatus and computer equipment | |
| JP2023502204A (en) | Neuromorphic device with crossbar array structure | |
| CN115204350A (en) | Training method and training device for impulse neural network | |
| CN117521741A (en) | Event-driven-based synaptic updating method and device | |
| CN116663622A (en) | Biological trusted neuron calculation circuit and calculation method | |
| CN113657592B (en) | Software-defined satellite self-adaptive pruning model compression method | |
| CN114925817A (en) | Data processing method, delay chain unit, delay device and many-core system | |
| CN111582461B (en) | Neural network training method and device, terminal equipment and readable storage medium | |
| US7117128B2 (en) | Method and apparatus for digital signal filtering | |
| Estévez et al. | Gamma SOM for temporal sequence processing | |
| CN113011572B (en) | Axon change amount determining method and device and weight processing method and device | |
| US12026623B2 (en) | Machine learning using gradient estimate determined using improved perturbations | |
| US20240346302A1 (en) | Hardware realization of neural networks using buffers | |
| Goryn et al. | Conjugate gradient learning algorithms for multilayer perceptrons | |
| EP3798919A1 (en) | Hardware architecture for spiking neural networks and method of operating |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |