Small-scale reserve pool calculation method and system based on photoelectric memristor

Abstract

The invention relates to a small-scale reserve pool calculation method and system based on a photoelectric memristor, and the photoelectric memristor is used as a basic processing unit to realize the integration of sensing, storage and calculation of optical information. Reservoir calculation of different optical signal sequences is realized by utilizing the volatile characteristic of the photoelectric memristor and the physical characteristic of different responses to the same excitation in different states. After different optical signals are applied to the device, the device shows different response states after being calculated by the storage pool so as to realize sensing and storage of the optical signals and calculation conversion of the optical signals to level signals. Therefore, a plurality of photoelectric memristors form a device array, a storage pool calculation network is established, the device array can be used for realizing classification and identification of optical sequence signals after training, and the network scale resource demand is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of micro-nano electronic devices, and in particular to a small-scale reserve pool calculation method and system based on photoelectric memristor. Background Art

[0002] The optical processing system based on the traditional von Neumann architecture is an architecture in which sensors, memory and processors are separated. The signal acquisition process needs to go through many intermediate processes such as analog-to-digital conversion, encoding, transmission and storage, and the calculation process data needs to be frequently exchanged between the memory and the processor, resulting in the generation of power consumption wall and storage wall. With the advent of the artificial intelligence era, the demand for optical information processing has exploded, posing a severe challenge to the current processing system.

[0003] Photoelectric memristor is a new type of micro-nano device that can change its resistance value by receiving stimulation from both optical and electrical signals. It also has synaptic plasticity and can simulate the biological retina to sense and pre-process optical information. This integrated sensing, storage and computing processing method is subversive compared to the traditional image processing split architecture and has important application prospects. However, the excitation signal of the photoelectric memristor has two dimensions, optical and electrical, and the modulation of the device characteristics is complex, and its pre-processing function still needs to be further developed.

[0004] Reservoir computing is a structural extension of recursive neural networks. It replaces the hidden layer of the original network with a storage pool unit with memory function, which can convert time series information into high-dimensional data for calculation, so that the network has the ability to process time series data. The characteristic of this network is that the weights inside the reserve pool remain unchanged, so only the weights of the linear fully connected output layer need to be updated, which can effectively reduce the difficulty and resource consumption of network training. At present, in the hardware implementation process of neural networks, how to use the physical properties of micro-nano devices to efficiently implement reserve pool units and effectively reduce the resource requirements of network scale is still a challenge in the industry. Summary of the invention

[0005] In view of the problems existing in the above-mentioned traditional methods, the present invention proposes a small-scale reserve pool calculation method based on photoelectric memristor and a small-scale reserve pool calculation system based on photoelectric memristor, which can effectively reduce the network scale resource requirements.

[0006] In order to achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:

[0007] On the one hand, a small-scale reservoir computing method based on a photoelectric memristor is provided, comprising:

[0008] Modulating the devices in the photoelectric memristor array of the reserve pool computing network to the same initialization conductance state; the reserve pool computing network includes an optical image receiving array and a back-end neural network, and the optical image receiving array is a photoelectric memristor array composed of multiple photoelectric memristors;

[0009] The optical image to be processed is converted into a multi-channel optical sequence signal according to four pixels as a group, and a multi-channel light source is used to generate a pulse sequence to simultaneously irradiate the optical sequence signal to the corresponding photoelectric memristor array; the number of photoelectric memristors in the optical image receiving array is the same as the number of channels, the photoelectric memristors correspond to the channels one by one, and each photoelectric memristor corresponds to four pixels of the optical image;

[0010] Read the state changes of the optical image receiving array, obtain the calculation results of the reserve pool and convert them into the input format required by the back-end neural network;

[0011] Input the calculation results of the reserve pool into the back-end neural network to complete the network operation of the current layer;

[0012] The output result of the current layer network operation is read as the classification and recognition result obtained by the current layer network operation on the optical sequence signal.

[0013] On the other hand, a small-scale reservoir computing system based on a photoelectric memristor is provided, comprising:

[0014] An initialization module is used to modulate the devices in the photoelectric memristor array of the reserve pool computing network to the same initialization conductance state; the reserve pool computing network includes an optical image receiving array and a back-end neural network, and the optical image receiving array is a photoelectric memristor array composed of multiple photoelectric memristors;

[0015] A signal input module is used to convert the optical image to be processed into a multi-channel optical sequence signal according to four pixels as a group, and use a multi-channel light source to generate a pulse sequence to simultaneously irradiate the optical sequence signal to the corresponding photoelectric memristor array; the number of photoelectric memristors in the optical image receiving array is the same as the number of channels, the photoelectric memristors correspond to the channels one by one, and each photoelectric memristor corresponds to four pixels of the optical image;

[0016] A state conversion module is used to read the state change of the optical image receiving array, obtain the calculation results of the reserve pool and convert them into the input format required by the back-end neural network;

[0017] The back-end input module is used to input the calculation results of the reserve pool into the back-end neural network to complete the current layer network calculation;

[0018] The result output module is used to read the output result of the current layer network operation as the classification and recognition result obtained by the current layer network operation on the optical sequence signal.

[0019] On the other hand, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned small-scale reserve pool calculation method based on photoelectric memristor are implemented.

[0020] On the other hand, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned small-scale reserve pool calculation method based on photoelectric memristor are implemented.

[0021] One of the above technical solutions has the following advantages and beneficial effects:

[0022] The above-mentioned small-scale reserve pool calculation method and system based on photoelectric memristor realizes the sensing, storage and calculation of optical information by adopting photoelectric memristor as the basic processing unit. The reserve pool calculation of different optical signal sequences is realized by utilizing the volatility of the photoelectric memristor itself and the physical characteristics of different responses to the same stimulus under different states. After different optical signals are applied to the device, the device exhibits different response states after the reserve pool calculation to realize the perception, storage and calculation conversion of the optical signal to the level signal. Accordingly, multiple photoelectric memristors are composed into a device array to establish a storage pool calculation network, which can be used to realize the classification and recognition of optical sequence signals after training, effectively reducing the network scale resource requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the conventional technology, the drawings required for use in the embodiments or the conventional technology descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 A schematic diagram of a flow chart of a small-scale reserve pool calculation method based on a photoelectric memristor in one embodiment;

[0025] Figure 2 A schematic diagram of a photoelectric memristor storage pool characteristic test in one embodiment;

[0026] Figure 3 A schematic diagram of the characteristic results of a photoelectric memristor storage pool in one embodiment;

[0027] Figure 4 A schematic diagram of a storage pool computing network structure based on a photoelectric memristor in one embodiment;

[0028] Figure 5 A schematic diagram showing a comparison of the scales of a reserve pool computing network and a traditional network when processing the same image in one embodiment;

[0029] Figure 6 Schematic diagram of the module structure of a small-scale reservoir computing system based on photoelectric memristor in one embodiment. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and Examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art of the present invention. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0031] It should be noted that the reference to "embodiment" in this article means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The presentation of this phrase at various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It will be appreciated by those skilled in the art that the embodiments described herein may be combined with other embodiments. The term "and / or" used in the specification of the present invention and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0032] The following will describe the implementation of the present invention in detail with reference to the accompanying drawings in the embodiment diagram of the present invention.

[0033] In one embodiment, Figure 1 As shown, a small-scale reserve pool calculation method based on a photoelectric memristor is provided, which may include the following processing steps S12 to S20:

[0034] S12, modulating the devices in the photoelectric memristor array of the reserve pool computing network to the same initialization conductance state; the reserve pool computing network includes an optical image receiving array and a back-end neural network, and the optical image receiving array is a photoelectric memristor array composed of multiple photoelectric memristors;

[0035] S14, converting the optical image to be processed into a multi-channel optical sequence signal according to four pixels as a group, using a multi-channel light source to generate a pulse sequence to simultaneously irradiate the optical sequence signal to the corresponding photoelectric memristor array; the number of photoelectric memristors in the optical image receiving array is the same as the number of channels, the photoelectric memristors correspond to the channels one by one, and each photoelectric memristor corresponds to four pixels of the optical image;

[0036] S16, reading the state change of the optical image receiving array, obtaining the calculation result of the reserve pool and converting it into the input format required by the back-end neural network;

[0037] S18, inputting the calculation results of the reserve pool into the back-end neural network to complete the current layer network calculation;

[0038] S20, reading the output result of the current layer network operation as the classification and recognition result obtained by the current layer network operation on the optical sequence signal.

[0039] It can be understood that the photoelectric memristor is a volatile device, including resistive, ferroelectric and phase change memristor devices. The optical excitation signal includes ultraviolet, infrared and visible light signals. The back-end connection of the photoelectric memristor is a neural network, including a single-layer or multi-layer fully connected network.

[0040] Photoelectric memristors can be used as basic processing units to realize the sensing, storage and calculation of optical information. By utilizing the volatility of photoelectric memristors and the different physical properties of the same stimulus under different states, the reserve pool calculation of different optical signal sequences can be realized. Apply 4-bit different optical signals to the photoelectric memristor, such as Figure 2 As shown in the figure, the device exhibits 16 different response states after calculation by the reserve pool, realizing the perception, storage and calculation conversion of optical signals into level signals, such as Figure 3 As shown. The basic network structure can be divided into a reserve pool computing layer, a neural network layer input layer, a hidden layer and an output layer. Based on the above characteristics of the photoelectric memristor, multiple photoelectric memristors can be combined into an optical image receiving array, such as Figure 4 The output of the reservoir array is transmitted to the fully connected neural network at the back end to form a reservoir computing network. After training, the back end network can classify and identify optical signals. Figure 4 The figure shows a schematic diagram of the network structure of the reserve pool calculation based on the photoelectric memristor. Input is the network input layer, Hide is the hidden layer, and Output is the network output layer. Figure 5 Same reason.

[0041] By utilizing the photoelectric memristor's ability to process light (timing) signals, the optical image is converted into a pulse sequence and input into the photoelectric memristor to form an optical image receiving array. Compared with the traditional optical sensor processing method, this can effectively reduce the scale of the sensor array and the scale requirements of the back-end neural network, such as Figure 5 As shown, when processing the same image, the size of the reserve pool computing network is only one-fourth of the traditional one.

[0042] Specifically, when the storage pool computing network is used to perform the classification and recognition task of the optical sequence signal, the following processing steps may be included:

[0043] The devices in the photoelectric memristor array of the reserve pool computing network are modulated to the same initialization conductance state using an existing resistance state modulation program; the same initialization conductance state can be a lower (such as the lowest) conductance state in the resistance change range of the photoelectric memristor.

[0044] The optical image to be processed is converted into a multi-channel optical sequence signal (i.e., a multi-channel pulse sequence) according to four pixels in a group, and then a multi-channel light source is used to generate a pulse sequence to simultaneously irradiate the optical sequence signal to the corresponding photoelectric memristor array. At this time, each photoelectric memristor senses the signal of its channel and performs existing reserve pool calculations. Depending on the light sequence signal, each photoelectric memristor can produce up to 16 different state changes; among them, the number of photoelectric memristors is the same as the number of channels, and the photoelectric memristors correspond to the channels one by one. Each photoelectric memristor corresponds to four pixels of the optical image and processes them into one data. Compared with the traditional one-to-one processing method of sensors and pixels, the number of photoelectric memristors is reduced to one-fourth, and the number of neurons in the input layer of the back-end neural network and the overall scale of the back-end neural network are also reduced to one-fourth.

[0045] The state changes of the photoelectric memristor array are read and converted into the back-end neural network input format; the back-end neural network can be efficiently implemented through neural network dedicated chips such as memristor chips, NPU (Neural Processing Unit) chips in the field, and its input signal can be a multi-channel voltage signal or a current signal.

[0046] Specifically, the state of the photoelectric memristor can be read through the existing analog integrated circuits in the art, for example, a small reading voltage is applied to each photoelectric memristor in the photoelectric memristor array, and the current value of the photoelectric memristor is detected, thereby realizing state reading. The state data format can be converted according to the input requirements of the hardware of the back-end neural network. For example, if the chip of the back-end neural network is current input, the read state change can be directly input into the back-end neural network, or the signal can be amplified by a multi-channel operational amplifier and then input into the back-end neural network; if the chip of the back-end neural network is voltage input, the read state change can be converted into a voltage signal by an existing multi-channel transconductance amplifier circuit and then input into the back-end neural network, realizing the conditioning and format conversion of the read state data.

[0047] The calculation results of the reserve pool are input into the back-end neural network to complete the current layer network operation; before actual use, the above-mentioned reserve pool calculation network can first select the data set corresponding to the scenario to be applied, such as the MNIST handwriting data set for network training, and use the existing back-propagation algorithm to adjust the network weights to obtain a trained reserve pool calculation network.

[0048] S20, reading the output result of the current layer network operation as the classification and recognition result obtained by the current layer network operation of the optical sequence signal; wherein, if the reserve pool computing network is a single-layer network, the classification and recognition of the optical sequence signal is completed after the result is obtained.

[0049] The above-mentioned small-scale reserve pool calculation method based on photoelectric memristors realizes the sensing, storage and calculation of optical information by using photoelectric memristors as basic processing units. The reserve pool calculation of different optical signal sequences is realized by utilizing the volatility of the photoelectric memristors themselves and the physical properties of different responses to the same stimulus under different states. After different optical signals are applied to the device, the device exhibits different response states after the reserve pool calculation to realize the perception, storage and calculation conversion of the optical signal to the level signal. Based on this, multiple photoelectric memristors are composed into a device array to establish a storage pool calculation network, which can be used to realize the classification and recognition of optical sequence signals after training, effectively reducing the network scale resource requirements.

[0050] Compared with traditional technologies, it realizes the perception and reserve pool calculation of optical sequence signals. This integrated sensing, storage and computing architecture avoids the delay and power consumption caused by analog-to-digital conversion and storage and computing separation, providing a disruptive change for optical data processing.

[0051] In one embodiment, when the reserve pool computing network is a multi-layer network structure, after the above step S20, the process jumps to step S14, and after performing all layer network operations to obtain the final output classification and recognition results, the classification and recognition of the optical sequence signal is completed.

[0052] It can be understood that if the reserve pool computing network has multiple layers of networks, the above steps S14 to S20 are continuously cycled to perform network operations on all layers, and the classification and recognition of the optical sequence signal is completed after the final output classification and recognition result is obtained.

[0053] The above-mentioned small-scale reservoir calculation method based on photoelectric memristor can be evaluated by combining experiments and simulations. First, the reservoir characteristics of the photoelectric memristor are tested by a semiconductor parameter analyzer and a probe station system to extract the parameter model. The mathematical model of the reservoir calculation network is established through simulation, and the measured parameters are substituted to select a suitable data set for training and simulation verification.

[0054] It should be understood that although Figure 1 The steps in the method are shown in sequence as indicated by the arrows, but the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of the steps, and the steps can be executed in other orders. Figure 1At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequentially, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0055] In one embodiment, Figure 6 As shown, a small-scale reserve pool computing system 100 based on photoelectric memristors is also provided, which may include an initialization module 11, a signal input module 13, a state conversion module 15, a back-end input module 17 and a result output module 19. Among them, the initialization module 11 is used to modulate the devices in the photoelectric memristor array of the reserve pool computing network to the same initialization conductance state; the reserve pool computing network includes an optical image receiving array and a back-end neural network, and the optical image receiving array is a photoelectric memristor array composed of multiple photoelectric memristors. The signal input module 13 is used to convert the optical image to be processed into a multi-channel optical sequence signal according to four pixels as a group, and a multi-channel light source is used to generate a pulse sequence to simultaneously irradiate the optical sequence signal to the corresponding photoelectric memristor array; the number of photoelectric memristors in the optical image receiving array is the same as the number of channels, and the photoelectric memristors correspond to the channels one by one, and each photoelectric memristor corresponds to four pixels of the optical image. The state conversion module 15 is used to read the state change of the optical image receiving array, obtain the reserve pool calculation result and convert it into the input format required by the back-end neural network. The back-end input module 17 is used to input the reserve pool calculation result into the back-end neural network to complete the current layer network operation. The result output module 19 is used to read the output result of the current layer network operation as the classification and recognition result obtained by the current layer network operation on the optical sequence signal.

[0056] The above-mentioned small-scale reserve pool computing system 100 based on photoelectric memristors realizes the sensing, storage and computing of optical information by using photoelectric memristors as basic processing units. The reserve pool computing of different optical signal sequences is realized by utilizing the volatility of the photoelectric memristors themselves and the physical properties of different responses to the same stimulus under different states. After different optical signals are applied to the device, the device exhibits different response states after the reserve pool computing to realize the perception, storage and computational conversion of optical signals to level signals. Accordingly, multiple photoelectric memristors are formed into a device array to establish a storage pool computing network, which can be used to realize the classification and recognition of optical sequence signals after training, effectively reducing the network scale resource requirements.

[0057] In one embodiment, when the reserve pool computing network is a multi-layer network structure, the result output module 19 jumps to the signal input module 13, performs all layer network operations to obtain the final output classification and recognition results, and completes the classification and recognition of the optical sequence signal.

[0058] Regarding the specific limitations of the small-scale reserve pool computing system 100 based on photoelectric memristors, please refer to the corresponding limitations of the small-scale reserve pool computing method based on photoelectric memristors above, which will not be repeated here. Each module in the above-mentioned small-scale reserve pool computing system 100 based on photoelectric memristors can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of a device with data processing functions in the form of hardware, or can be stored in the memory of the aforementioned device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. The aforementioned device can be, but is not limited to, various types of data computer devices already available in the art.

[0059] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the following processing steps when executing the computer program: modulating devices in a photoelectric memristor array of a reserve pool computing network to the same initialization conductance state; the reserve pool computing network includes an optical image receiving array and a back-end neural network, and the optical image receiving array is a photoelectric memristor array composed of multiple photoelectric memristors; converting an optical image to be processed into a multi-channel optical sequence signal according to four pixels as a group, and using a multi-channel light source to generate a pulse sequence to simultaneously irradiate the optical sequence signal to the corresponding photoelectric memristor array; the number of photoelectric memristors in the optical image receiving array is the same as the number of channels, the photoelectric memristors correspond to the channels one by one, and each photoelectric memristor corresponds to four pixels of the optical image; reading the state change of the optical image receiving array, obtaining the reserve pool calculation result and converting it into the input format required by the back-end neural network; inputting the reserve pool calculation result into the back-end neural network to complete the current layer network operation; reading the output result of the current layer network operation as the classification and recognition result obtained by the current layer network operation of the optical sequence signal.

[0060] In one embodiment, when the processor executes the computer program, the added steps or sub-steps in each embodiment of the above-mentioned small-scale reserve pool calculation method based on photoelectric memristor can also be implemented.

[0061] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following processing steps are implemented: devices in a photoelectric memristor array of a reserve pool computing network are modulated to the same initialization conductance state; the reserve pool computing network includes an optical image receiving array and a back-end neural network, and the optical image receiving array is a photoelectric memristor array composed of multiple photoelectric memristors; the optical image to be processed is converted into a multi-channel optical sequence signal according to four pixels as a group, and a multi-channel light source is used to generate a pulse sequence to simultaneously irradiate the optical sequence signal to the corresponding photoelectric memristor array; the number of photoelectric memristors in the optical image receiving array is the same as the number of channels, and the photoelectric memristors correspond to the channels one by one, and each photoelectric memristor corresponds to four pixels of the optical image; the state change of the optical image receiving array is read to obtain the reserve pool calculation result and convert it into the input format required by the back-end neural network; the reserve pool calculation result is input into the back-end neural network to complete the current layer network operation; the output result of the current layer network operation is read as the classification and recognition result obtained by the current layer network operation of the optical sequence signal.

[0062] In one embodiment, when the computer program is executed by a processor, it can also implement the steps or sub-steps added in each embodiment of the above-mentioned small-scale reserve pool calculation method based on photoelectric memristor.

[0063] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus dynamic random access memory (RambusDRAM, referred to as RDRAM) and interface dynamic random access memory (DRDRAM).

[0064] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0065] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of protection of the invention. It should be pointed out that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the attached claims.

Claims

1. A small-scale reservoir computing method based on photoelectric memristor, It is characterized in that Includes steps: Modulating the devices in the photoelectric memristor array of the reserve pool computing network to the same initialization conductance state; the reserve pool computing network includes an optical image receiving array and a back-end neural network, and the optical image receiving array is a photoelectric memristor array composed of multiple photoelectric memristors; The optical image to be processed is converted into a multi-channel optical sequence signal according to four pixels in a group, and a multi-channel light source is used to generate a pulse sequence to simultaneously irradiate the optical sequence signal to the corresponding photoelectric memristor array; The number of photoelectric memristors in the optical image receiving array is the same as the number of channels, the photoelectric memristors correspond to the channels one by one, and each photoelectric memristor corresponds to four pixels of the optical image; Reading the state change of the optical image receiving array, obtaining the calculation result of the reserve pool and converting it into the input format required by the back-end neural network; Input the calculation results of the reserve pool into the back-end neural network to complete the current layer network operation; The output result of the current layer network operation is read as the classification and recognition result obtained by the current layer network operation on the optical sequence signal.

2. The small-scale reservoir calculation method based on photoelectric memristor according to claim 1, It is characterized in that When the reserve pool computing network is a multi-layer network structure, after the step of reading the output result of the current layer network operation as the classification and recognition result obtained by the current layer network operation of the optical sequence signal, jump to the step of converting the optical image to be processed into a multi-channel optical sequence signal according to four pixels as a group, using a multi-channel light source to generate a pulse sequence to simultaneously irradiate the optical sequence signal to the corresponding photoelectric memristor array, and after performing all layer network operations to obtain the final output classification and recognition result, the classification and recognition of the optical sequence signal is completed.

3. A small-scale reservoir computing system based on photoelectric memristors, It is characterized in that include: An initialization module, used for modulating the devices in the photoelectric memristor array of the reserve pool computing network to the same initialization conductance state; The reserve pool computing network includes an optical image receiving array and a back-end neural network, wherein the optical image receiving array is a photoelectric memristor array composed of a plurality of photoelectric memristors; A signal input module, used to convert the optical image to be processed into a multi-channel optical sequence signal according to four pixels as a group, and use a multi-channel light source to generate a pulse sequence to simultaneously irradiate the optical sequence signal to the corresponding photoelectric memristor array; The number of photoelectric memristors in the optical image receiving array is the same as the number of channels, the photoelectric memristors correspond to the channels one by one, and each photoelectric memristor corresponds to four pixels of the optical image; A state conversion module, used to read the state change of the optical image receiving array, obtain the calculation result of the reserve pool and convert it into the input format required by the back-end neural network; The back-end input module is used to input the calculation results of the reserve pool into the back-end neural network to complete the current layer network calculation; The result output module is used to read the output result of the current layer network operation as the classification and recognition result obtained by the current layer network operation on the optical sequence signal.

4. The small-scale reservoir computing system based on photoelectric memristor according to claim 3, It is characterized in that When the reserve pool computing network is a multi-layer network structure, the result output module jumps to the signal input module, performs all layer network operations to obtain the final output classification and recognition results, and completes the classification and recognition of the optical sequence signal.

5. A computer device comprising a memory and a processor, wherein the memory stores a computer program, It is characterized in that When the processor executes the computer program, the steps of the small-scale reservoir calculation method based on photoelectric memristor according to claim 1 or 2 are implemented.

6. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the small-scale reservoir calculation method based on photoelectric memristor according to claim 1 or 2 are implemented.

Images