Micro-nano material physical reserve pool with mixed topological structure and corresponding computing system

Abstract

The invention belongs to the field of novel computing hardware, and discloses a micro-nano material physical reserve pool of a hybrid topological structure and a corresponding computing system.The micro-nano material physical reserve pool of the hybrid topological structure comprises a connecting plate, a first micro-nano material network physical reserve pool and a second micro-nano material network physical reserve pool, and the first micro-nano material network physical reserve pool and the second micro-nano material network physical reserve pool are located on the front face and the back face of the connecting plate respectively; the two micro-nano material network physical reserve pools are independently selected from ESN and LSM; the nodes of the two networks are stacked or connected in parallel. The first micro-nano material network physical reserve pool and the second micro-nano material network physical reserve pool are constructed on the front face and the back face of the same connecting plate respectively, so that the physical reserve pool of a mixed topological structure is formed, the nonlinear mapping capacity and the parallel processing capacity of the system can be improved, and the training complexity is reduced; the technical problems that an existing carbon nano tube physical reserve pool is low in calculation performance and limited in scale can be solved.

Description

Technical Field

[0001] The present invention belongs to the field of novel computing hardware, and more specifically, relates to a physical reservoir of micro-nano materials with a hybrid topology and a corresponding computing system. Background Art

[0002] Reservoir Computing is a computing architecture derived from recurrent neural networks, which has the advantages of low training cost and fast learning. The internal connection weights of the reservoir are randomly generated and do not need to be adjusted, so they can be replaced by physical systems such as new materials and devices with nonlinear dynamics. The network physical reservoir based on micro-nano materials is an emerging type among them.

[0003] Among micro-nano materials, carbon nanotubes (CNTs) are a new type of nano material with a unique structure, having significant advantages such as ultra-small size, high mobility, high saturation velocity, and long mean free path. The reservoir system based on carbon nanotubes is also one of the research hotspots of physical reservoirs.

[0004] Among them, the research on the 2D single-walled carbon nanotube (SWCNT) network physical reservoir is relatively extensive. From 2016 to 2017, the Matthew Dale team used SWCNT and polymers to form a reservoir, using two polymers, polymethyl methacrylate (PMMA) and polybutyl methacrylate (PBMA), and pre-trained the SWCNT / polymer device using an evolutionary algorithm, configuring it into a functional RC system. This device completed tasks such as nonlinear sequence fitting, waveform generation, and MC calculation, and performed better than the silver nanowire network reservoir, demonstrating the advantages of carbon nanotubes as a reservoir computing base material; however, due to the kinetic characteristics of the conductive wire network used in the SWCNT / polymer reservoir, the nonlinear response to input signals is insufficient, the internal nonlinear dynamics of the network are not rich enough, and the overall computing performance of the reservoir is relatively average.

[0005] In 2022, the Megumi Akai-Kasaya team built a POM / SWCNT random reservoir network by using a composite structure formed by adsorbing carbon nanotubes with phosphomolybdic acid (POM), and achieved basic reservoir computing, such as periodic sequence prediction tasks, NARMA2, and MC calculation. Among them, the functionalization of SWCNT by POM makes the composite structure have more complex and rich nonlinear electrical characteristics.

[0006] Due to the low interconnection probability of carbon nanotubes inside the 2D SWCNT network physical reservoir, the 2D reservoir has problems such as limited scale and limited complexity of connections between nodes, resulting in limited overall nonlinearity of the 2D reservoir and low overall computing performance of the reservoir.

[0007] Physical reservoir pools constructed with other micro-nano materials also have similar deficiencies. Summary of the Invention

[0008] In view of the above deficiencies or improvement requirements of the prior art, the purpose of the present invention is to provide a physical reservoir pool of micro-nano materials with a hybrid topological structure and a corresponding computing system. By constructing a first micro-nano material network physical reservoir pool (such as a first carbon nanotube network physical reservoir pool) and a second micro-nano material network physical reservoir pool (such as a second carbon nanotube network physical reservoir pool) on the front and back sides of the same connection board (such as a PCB board), a physical reservoir pool with a hybrid topological structure is formed, which can improve the non-linear mapping ability and parallel processing ability of the system and reduce the complexity of training. Taking a carbon nanotube network with carbon nanotubes as micro-nano materials as an example, based on the present invention, two carbon nanotube networks (such as two 2D carbon nanotube networks) are assembled to form a physical reservoir pool with a hybrid topological structure, and the obtained hybrid topological structure can solve the technical problems of low computing performance and limited scale of the existing carbon nanotube physical reservoir pool.

[0009] To achieve the above object, according to one aspect of the present invention, there is provided a physical reservoir pool of micro-nano materials with a hybrid topological structure, including a connection board, a first micro-nano material network physical reservoir pool located on the front side of the connection board, and a second micro-nano material network physical reservoir pool located on the back side of the connection board; wherein, the first micro-nano material network physical reservoir pool and the second micro-nano material network physical reservoir pool are independently selected from: a micro-nano material network physical reservoir pool based on an Echo State Network (ESN), a micro-nano material network physical reservoir pool based on a Liquid State Machine (LSM);

[0010] The input nodes of the first micro-nano material network physical reservoir pool are connected in parallel with the input nodes of the second micro-nano material network physical reservoir pool, and the output nodes of the first micro-nano material network physical reservoir pool are connected in parallel with the output nodes of the second micro-nano material network physical reservoir pool; or, the input nodes of the first micro-nano material network physical reservoir pool are short-circuited one by one with the input nodes of the second micro-nano material network physical reservoir pool, and the output nodes of the first micro-nano material network physical reservoir pool are short-circuited one by one with the output nodes of the second micro-nano material network physical reservoir pool.

[0011] As a further preference of the present invention, the micro-nano material is a carbon nanotube, and both the first micro-nano material network physical reservoir pool and the second micro-nano material network physical reservoir pool are carbon nanotube network physical reservoir pools.

[0012] As a further preference of the present invention, both the first micro-nano material network physical reservoir pool and the second micro-nano material network physical reservoir pool are 2D carbon nanotube network physical reservoir pools.

[0013] As a further preference of the present invention, the reservoir of the first micro-nano material network physical reservoir is directly prepared on the connection board, or is prepared on a silicon substrate and then mounted on the connection board;

[0014] The reservoir of the second micro-nano material network physical reservoir is directly prepared on the connection board, or is prepared on a silicon substrate and then mounted on the connection board.

[0015] As a further preference of the present invention, the connection board is a PCB board.

[0016] As a further preference of the present invention, the carbon nanotubes are semiconducting single-walled carbon nanotubes.

[0017] As a further preference of the present invention, the micro-nano material network physical reservoir based on the Echo State Network (ESN) is the carbon nanotube network physical reservoir based on the Echo State Network (ESN), specifically a POM / SWCNT / PBMA thin film device formed by components including phosphomolybdic acid, single-walled carbon nanotubes, and polybutyl methacrylate;

[0018] The micro-nano material network physical reservoir based on the Liquid State Machine (LSM) is the carbon nanotube network physical reservoir based on the Liquid State Machine (LSM), specifically a POM / SWCNT pulse device formed by components including phosphomolybdic acid and single-walled carbon nanotubes.

[0019] As a further preference of the present invention, one of the first micro-nano material network physical reservoir and the second micro-nano material network physical reservoir is a micro-nano material network physical reservoir based on the Echo State Network (ESN), and the other is a micro-nano material network physical reservoir based on the Liquid State Machine (LSM).

[0020] According to another aspect of the present invention, the present invention provides a computing system including the micro-nano material physical reservoir with the above-mentioned hybrid topology structure.

[0021] Through the above technical solution conceived by the present invention, compared with the prior art, based on the physical reservoir with the hybrid topology structure in the present invention, the nodes of the two networks are stacked or connected in parallel to respectively realize a composite node or a reservoir with an enlarged scale. Currently, the research on reservoir computing mainly focuses on the theory and model, and the device implementation of the physical reservoir also mostly focuses on the implementation and research of a single network. Based on the implementation of individual ESN and LSM network devices, considering the high demand of the computing performance of the reservoir for the nonlinearity of the system, the nonlinear characteristics and reservoir computing performance of the system are enhanced by topologically mixing multiple network devices.

[0022] (a) Taking the hybridization of ESN networks as an example, the capacity of the model can be expanded, enabling it to capture and represent more complex dynamic relationships and features. In some complex prediction tasks, a single ESN may be difficult to achieve the desired accuracy. By combining multiple ESNs, the synergistic effect between them can be utilized to improve the prediction accuracy of complex systems and also enable more flexible adaptation to various different application scenarios.

[0023] Different from simply increasing the number of nodes in the same ESN network, although increasing the number of nodes can improve the fitting ability of the network and thus better capture the complex dynamic patterns in the data, the disadvantages brought about by too many nodes cannot be ignored. For example, as the node scale expands, the computational complexity and storage requirements will also increase correspondingly, resulting in low computational efficiency. Too many nodes may lead to an overly complex network, prone to overfitting problems and reduced generalization ability. Increasing the number of nodes will increase the difficulty of adjusting parameters such as connection weights, output weights, and spectral radius. In addition, simply relying on increasing the number of nodes to enhance the computational ability of the physical reservoir does not meet the development requirements of electronic process integration. The present invention, however, utilizes the hybridization of ESN networks, which can improve computational efficiency and enhance the model's ability to capture complex dynamic patterns, while reducing storage requirements and the risk of overfitting by optimizing the network structure. Similar advantages also hold for the hybridization of LSM networks and the hybridization of ESN networks and LSM networks.

[0024] (b) Hybridizing LSM networks further enhances the system's parallel processing ability, enabling it to handle a large amount of spatio-temporal information simultaneously. At the same time, it can also expand the scope and depth of the system's recognition of complex spatio-temporal patterns, making it perform better in tasks such as processing complex biological signals and speech recognition. The combination of multiple LSMs can form a more robust network structure, and even if some LSMs are interfered with or damaged, the entire system can still maintain good performance.

[0025] (c) Especially when one of the first micro-nano material network physical reservoirs and the second micro-nano material network physical reservoirs is a micro-nano material network physical reservoir based on an echo state network (ESN), and the other is a micro-nano material network physical reservoir based on a liquid state machine (LSM):

[0026] (1) The present invention parallels the ESN network structure and the LSM network structure. ESN has good short-term memory ability, and LSM is suitable for long-term dependence and processing of complex dynamics. Hybridizing the two can form a comprehensive response to different features in the time series, enhancing the system's time response ability.

[0027] (2) The liquid layer of LSM topologically simulates the processing mode of biological neural networks, with strong nonlinearity and noise robustness. After mixing, it can make up for the deficiencies of ESN in a high-noise environment and enhance the nonlinear characteristics of the system;

[0028] (3) By mixing the two network structures, more diverse dynamic states can be formed in the reservoir, enabling the output layer to make more accurate predictions based on richer state information and improving the computational performance of the physical reservoir of micro-nano materials. Brief Description of the Drawings

[0029] Figure 1 It is a schematic diagram of assembling the ESN network on one side of the PCB board provided by the present invention.

[0030] Figure 2 It is a schematic diagram of assembling the LSM network on one side of the PCB board provided by the present invention.

[0031] Figure 3 It is a schematic diagram of the stacked structure of the ESN network and the LSM network provided by the present invention.

[0032] Figure 4 It is a schematic diagram of the parallel structure of the ESN network and the LSM network provided by the present invention.

[0033] Figure 5 It is a physical diagram of the silicon-based POM / CNT / PBMA thin-film device serving as the physical reservoir of the ESN network provided by the present invention.

[0034] Figure 6 It is a physical diagram of the silicon-based POM / SWCNT pulse device serving as the physical reservoir of the LSM network provided by the present invention.

[0035] Figure 7 It is a schematic diagram of the structure of the reservoir computing test system provided by the present invention.

[0036] Figure 8 It is a schematic diagram of the result of using a single silicon-based ESN network device to perform the NARMA10 time series prediction task in Example 1.

[0037] Figure 9 It is a schematic diagram of the result of using a single silicon-based LSM network device to perform the NARMA10 time series prediction task in Example 1.

[0038] Figure 10 It is a schematic diagram of the result of using the silicon-based ESN network device and the LSM network device in parallel to perform the NARMA10 time series prediction task provided in Example 1 of the present invention.

[0039] Figure 11It is a schematic diagram of the result of using two stacked silicon-based ESN network devices for the NARMA10 time series prediction task provided in Embodiment 2 of the present invention. Detailed implementation manners

[0040] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0041] Taking a 2D carbon nanotube network and a PCB board as the connection board as an example, in the following embodiments, the preparation methods of the ESN network and the LSM network are as follows (the carbon nanotubes used are specifically commercially available semiconducting single-walled carbon nanotubes):

[0042] The topological structure of the physical reservoir of the ESN network is as Figure 1 shown. The microelectrode array serving as the connection node can be directly designed on the PCB board or a metal thin film electrode can be prepared on a silicon substrate through a patterning process, and then the silicon wafer is mounted on the PCB board; among them, the main process of the patterning process is: forming a 10*10 microelectrode array pattern on the silicon wafer surface through photolithography, sputtering to obtain nickel electrode thin films with a thickness of 50 nm and gold electrode thin films with a thickness of 350 nm respectively and annealing in vacuum at 400 °C for 4 hours, preparing a POM / SWCNT / PBMA anisole solution, where the SWCNT concentration is 500 μg / ml, the mass ratio of SWCNT to PBMA is 1%, the concentration range of POM is 200 - 800 μg / ml, and the concentration used here is 500 μg / ml. Finally, a POM / SWCNT / PBMA network thin film is generated on the microelectrode array surface by dropping and drying at 100 °C. Finally, the electrodes on the silicon wafer are connected to the corresponding positions on the PCB board through wire bonding to obtain a microelectrode array that can be used for testing.

[0043] The CNTs in the POM / SWCNT / PBMA network thin film randomly interconnect and fixedly connect the nodes of the electrode array, constituting the first-layer network topology of the physical reservoir of the ESN network, having the characteristics of randomly initializing and fixedly maintaining the connection weights; the non-repeatability between the thin films prepared by the dropping process improves the complexity inside the network; the nodes inside the thin film and the input and output nodes on the PCB board constitute the second-layer topology.

[0044] The topological structure of the physical reservoir of the LSM network is as Figure 2As shown, similar to the ESN network, a microelectrode array on a silicon substrate or a PCB board can be selected as needed. The main process for fabricating the device on a silicon wafer is as follows: A 16×2 microelectrode array pattern is formed on the silicon wafer surface through photolithography. The electrode spacing for building CNTs is 2 μm. Nickel electrode films with a thickness of 35 nm and gold electrode films with a thickness of 500 nm are sputtered and vacuum annealed at 400 °C for 4 hours. A POM / SWCNT ethanol solution is prepared by ultrasonic mixing, where the concentrations of POM and SWCNT are 300 μg / ml and 20 μg / ml, respectively. CNTs are randomly assembled between the electrode arrays on the surface of the microelectrode array through dielectrophoresis (DEP) to form a path. Finally, the electrodes on the silicon wafer are connected to the corresponding ones on the PCB board through wire bonding to obtain a microelectrode array for testing.

[0045] During electrophoresis, the number of CNTs is related to DEP parameters such as the signal voltage amplitude, frequency, solution concentration, and assembly time. In the present invention, the signal voltage amplitude is set to 16 V during DEP pp , and the frequency is 2 MHz. A randomly connected sparse network is generated between planar electrodes through DEP, forming the network topology inside the physical reservoir of the LSM network, which has a high degree of nonlinearity and complexity.

[0046] Example 1

[0047] In this example, the ESN network is assembled on the front side of the same PCB board, and the LSM network is assembled on the back side, as shown in Figure 1 、 Figure 2 .

[0048] When the corresponding device is in operation, it can be either in a stacked manner or in a parallel manner. Specifically:

[0049] Figure 3 is one of the hybrid topological structures provided by the present invention for stacking the ESN network and the LSM network. The physical reservoirs of the ESN network with M nodes and the LSM network with M nodes are respectively assembled on the two sides of the PCB board. The input and output nodes of the two physical reservoirs are short-circuited respectively, that is, M input nodes simultaneously provide input signals to the corresponding network nodes on both sides. After being processed separately inside the reservoirs, the output nodes simultaneously collect the M output states of the two networks. Since the connection nodes inside the LSM network are in one-to-one correspondence, while the connection nodes of the ESN network are randomly distributed, it results in that the response collected by the same output node may come from two or more input nodes. The output layer can collect richer state information. When a certain node has a problem, other nodes can still continue to work, improving the complexity and fault tolerance of the system.

[0050] Figure 4Another hybrid topology provided by the present invention is to parallelize the ESN network and the LSM network. The physical reservoir of the ESN network with M nodes and the physical reservoir of the LSM network with N nodes are respectively assembled on both sides of the PCB board. The (M + N) input nodes simultaneously provide input signals. After being processed separately in the reservoir, (M + N) output states are obtained. By paralleling the two reservoirs, the scale of the state space is expanded, and the processing ability of the system for complex inputs and the storage ability for temporal information are enhanced, thereby improving the computing performance of reservoir computing.

[0051] The silicon-based devices of the ESN and LSM physical reservoirs are as Figure 5 and Figure 6 shown. The silicon wafer is mounted on the PCB, and the electrodes of the silicon-based devices are connected to the electrodes of the PCB board using the wire bonding process. Then, it is connected to the upper computer software system through a hardware system including an industrial computer, an NI data acquisition card, etc. to perform signal input, acquisition, and reservoir computing task testing, constituting a complete physical reservoir computing system, as Figure 7 shown (the reservoir layer / liquid layer is the above-mentioned SWNT network device). Based on this computing system, the Nonlinear Auto Regressive Moving Average (NARMA) time series prediction task is respectively performed using independent ESN network devices and LSM network devices. Since NARMA10 has a higher order than NARMA2 (NARMA10 is a tenth-order nonlinear model, and its model is more complex because more historical information needs to be considered; NARMA2 is a second-order nonlinear model, and its model is relatively simple and has a lower computational complexity due to the lower order), it can better test the performance and memory depth of the system when dealing with complex nonlinear dynamic systems. When verifying the performance of the network devices in the present invention, the NARMA10 task is used for calculation, and the results are as Figure 8 、 Figure 9 shown. The physical reservoir network devices provided by the present invention can better complete the time series prediction task, and the normalized mean square error values (NMSE) are 0.027647 and 0.061098 respectively.

[0052] Figure 10To calculate the results of the NARMA10 task after paralleling the ESN network device and the LSM network device, the NMSE is 0.016605, which is more accurate than the prediction results of the independent ESN / LSM network. It can be seen that the computing performance of the hybrid network obtained based on the present invention is superior to that of a single network. The structure of the physical reservoir of the carbon nanotube network with the hybrid topology structure of the present invention can further enhance the internal complexity of the network, strengthen the nonlinear characteristics and dynamic characteristics of the entire network system by mixing the above two network topology structures, thereby improving the computing performance of the carbon nanotube physical reservoir. This is not only applicable to carbon nanotube networks, but also to the physical reservoirs of micro-nano material networks formed by other micro-nano materials.

[0053] Example 2

[0054] In addition to the above Example 1 Figure 3 and Figure 4 the topology structure of mixing the ESN network and the LSM network shown, the hybrid network topology structure can also adopt other methods. In this Example 2, the ESN network is assembled on the front side of the same PCB board, and the ESN network is also assembled on the back side (that is, two ESN networks are assembled on the same PCB board at the same time). Between these two networks, they can also be stacked or paralleled.

[0055] Figure 11 To calculate the results of the NARMA10 task after stacking two ESN network devices, the NMSE is 0.016566, which is more accurate than the prediction results of the independent ESN network. It can be seen that the computing performance of the hybrid network obtained based on the present invention is superior to that of a single network. Mixing the ESN network and the ESN network can expand the capacity of the model, enabling it to capture and represent more complex dynamic relationships and features; in some complex prediction tasks, a single ESN may be difficult to achieve ideal accuracy. By combining multiple ESNs, the synergy between them can be utilized to improve the prediction accuracy of complex systems and can also more flexibly adapt to various different application scenarios.

[0056] Example 3

[0057] In this Example 3, the LSM network is assembled on the front side of the same PCB board, and the LSM network is also assembled on the back side (that is, two LSM networks are assembled on the same PCB board at the same time). Between these two networks, they can also be stacked or paralleled.

[0058] Mixing the LSM network and the LSM network can further enhance the parallel processing ability of the system, enabling it to process a large amount of spatio-temporal information simultaneously; at the same time, it can also expand the recognition range and depth of the system for complex spatio-temporal patterns, making it perform better in processing tasks such as complex biological signals and speech recognition; the combination of multiple LSMs can form a more robust network structure, and even if some LSMs are interfered with or damaged, the entire system can still maintain good performance.

[0059] The above embodiments are only examples. The same or different types of physical reserve pools can be assembled on the front and back sides of the PCB. That is to say, whether to construct a topological structure that mixes the ESN network and the LSM network, or a topological structure that mixes two ESN networks, or a topological structure that mixes two LSM networks, can be selected according to actual computing requirements. In addition to the PCB board, other connection boards can also be used for the connection board. In addition, the above embodiments provide a structural design for assembling two 2D carbon nanotube networks. In addition to the 2D carbon nanotube network, networks formed by other micro-nano materials can also be used (of course, it can also be a 3D network device).

[0060] For the parts not detailed in the present invention, reference can be made to the prior art. For example, the stacking or parallel connection of the nodes of the two networks (corresponding to the connection between the input layers of the two networks and the connection between the output layers of the two networks) are all realized by the software system and the hardware system (for example, the hardware system mainly consists of an industrial control computer, an NI data acquisition card, and an input / output control circuit board for the acquisition interface; the software system is mainly the upper computer program on the LabVIEW and MATLAB platforms).

[0061] Those skilled in the art can easily understand that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A micro-nano material physical storage pool with a mixed topological structure, characterized in that: The invention comprises a connecting plate, a first micro-nano material network physical reservoir located on the front side of the connecting plate, and a second micro-nano material network physical reservoir located on the back side of the connecting plate; wherein the first micro-nano material network physical reservoir and the second micro-nano material network physical reservoir are independently selected from: a micro-nano material network physical reservoir based on an echo state network (ESN) and a micro-nano material network physical reservoir based on a liquid state machine (LSM); The input node of the first micro-nano material network physical reserve pool is connected in parallel with the input node of the second micro-nano material network physical reserve pool, and the output node of the first micro-nano material network physical reserve pool is connected in parallel with the output node of the second micro-nano material network physical reserve pool; or, the input node of the first micro-nano material network physical reserve pool is short-circuited with the input node of the second micro-nano material network physical reserve pool, and the output node of the first micro-nano material network physical reserve pool is short-circuited with the output node of the second micro-nano material network physical reserve pool.

2. The micro-nano material physical storage pool of the hybrid topological structure as claimed in claim 1, characterized in that: The micro-nano material is a carbon nanotube, and the first micro-nano material network physical reserve pool and the second micro-nano material network physical reserve pool are both carbon nanotube network physical reserve pools.

3. The micro-nano material physical storage pool of the hybrid topological structure as claimed in claim 2, characterized in that: The first micro-nano material network physical reserve pool and the second micro-nano material network physical reserve pool are both 2D carbon nanotube network physical reserve pools.

4. The micro-nano material physical storage pool of the hybrid topological structure as claimed in claim 1, characterized in that: The reservoir of the first micro-nano material network physical reservoir is directly prepared on the connecting plate, or is prepared on a silicon substrate and then mounted on the connecting plate; The reservoir of the second micro-nano material network physical reservoir is directly prepared on the connecting board, or is prepared on a silicon substrate and then mounted on the connecting board.

5. The micro-nano material physical storage pool of hybrid topological structure as claimed in claim 4, characterized in that: The connecting board is a PCB board.

6. The micro-nano material physical storage pool of mixed topological structure as claimed in claim 2, characterized in that: The carbon nanotubes are semiconducting single-walled carbon nanotubes.

7. The micro-nano material physical storage pool of hybrid topological structure as claimed in claim 1, characterized in that: The micro-nano material network physical reserve pool based on echo state network (ESN) is the carbon nanotube network physical reserve pool based on echo state network (ESN), specifically a POM / SWCNT / PBMA thin film device formed by components including phosphomolybdic acid, single-walled carbon nanotubes and polybutyl methacrylate; The micro-nano material network physical reserve pool based on liquid state machine (LSM) is the carbon nanotube network physical reserve pool based on liquid state machine (LSM), specifically a POM / SWCNT pulse device formed by phosphomolybdic acid and single-walled carbon nanotubes.

8. The micro-nano material physical storage pool of mixed topological structure as claimed in claim 1, characterized in that: One of the first micro-nano material network physical reservoir and the second micro-nano material network physical reservoir is a micro-nano material network physical reservoir based on echo state network (ESN), and the other is a micro-nano material network physical reservoir based on liquid state machine (LSM).

9. A computing system, characterized in that: A physical storage pool of micro-nano materials comprising a hybrid topological structure as claimed in any one of claims 1 to 8.

Images