Spin-torque nano-oscillator based reservoir computing systems, methods, and media

Abstract

The present application relates to a spin-torque nano-oscillator-based reservoir computing system, method and medium, belonging to the field of neuromorphic computing and hardware acceleration, which directly generates nonlinear signals from linear signals by using the inherent nonlinear current-voltage characteristics of spin-torque nano-oscillators, realizes the nonlinear part generation of feature vectors in the next generation of reservoir computing, and effectively solves the three major problems of feature dimension combination explosion, high computational overhead caused by a large number of matrix operations, and insufficient multi-scale dynamics capture capability by replacing traditional polynomial algorithms with physical laws. At the hardware level, adaptive frequency domain response is realized without complex reconfiguration, combined with low computational overhead of analog domain processing, which can improve the timing processing efficiency without significantly increasing the hardware cost.

Description

Technical Field

[0001] This invention belongs to the field of neuromorphic computing and hardware acceleration technology, and relates to a reservoir computing system, method and medium based on a spin torque nano-oscillator. Background Technology

[0002] The spin-torque nano-oscillator (STNO) based on a magnetic tunnel junction (MTJ) is a magnetic device consisting of an ultrathin barrier layer and two ferromagnetic (FM) layers. One FM layer has a fixed magnetic orientation (fixed layer / reference layer), while the magnetic orientation of the other FM layer (free layer) can be changed. The core working mechanism of the STNO is based on the physical foundation of the precise coupling between spin transfer torque (STT) and magnetodynamics. When the injected current intensity exceeds a critical threshold, the spin transfer torque effect precisely cancels the magnetic damping force of the free layer, forcing the magnetization vector to leave static equilibrium and enter a stable continuous precession state. This continuous magnetization oscillation is converted into a measurable voltage oscillation signal in real time through the magnetoresistance effect, and its amplitude directly maps the intensity of the magnetization precession. Furthermore, the voltage amplitude, as a function of the injected current, is highly nonlinear and essentially depends on past inputs. By utilizing the amplitude dynamics of the STNO, the nonlinearity and memory characteristics of neuromorphic computing are combined in a single nanodevice.

[0003] Reservoir computing (RC) is a computational framework originating from recurrent neural networks (RNNs) specifically designed for processing time-series data. It is widely used in tasks requiring contextual understanding, such as natural language processing and time-series analysis. In RC, the low-dimensional input signal is first mapped to a high-dimensional dynamical system called the reservoir, typically composed of an RNN with a fixed random input matrix and internal connection matrices. After nonlinear transformation by the reservoir, the input data is converted into high-dimensional spatiotemporal dynamic features, which are then decoded through a lightweight, trainable output layer. The core structure of traditional RC involves randomly generated key parameters, lacking explicit design rules, leading to uncertainties in the results.

[0004] Research on Next Generation Reservoir Computing (NG-RC) has theoretically demonstrated that reservoir computing with linear reservoir nodes and a nonlinear output layer is mathematically equivalent to a nonlinear vector autoregression (NVAR) method. This discovery means that powerful reservoir computing capabilities can be achieved without actually building and running a complex stochastic RNN (i.e., a "reservoir"). The NVAR method directly utilizes the time delay and nonlinear combination of the time series data itself to construct feature vectors, thus implicitly defining an equivalent reservoir computing, requiring far fewer parameters in the output weight matrix to be trained compared to traditional reservoir computing. However, current designs of NG-RC still suffer from low performance in time series signal processing. Summary of the Invention

[0005] To address the problems existing in the above-mentioned traditional methods, this invention proposes a reservoir calculation method based on a spin torque nano-oscillator, a reservoir calculation system based on a spin torque nano-oscillator, and a computer-readable storage medium, which can realize efficient and adaptive time-series signal processing.

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

[0007] On the one hand, a reservoir computing system based on a spin-torque nano-oscillator is provided. The spin-torque nano-oscillator serves as a nonlinear node in next-generation reservoir computing, used to generate the nonlinear part of the eigenvector in next-generation reservoir computing. This reservoir computing system includes:

[0008] The data expansion module is used to expand the external data to be processed into a linear part of a one-dimensional feature vector based on the dimension and time sequence relationship of the external data to be processed.

[0009] The injection oscillation module is used to convert the linear part into an input current and inject it into the spin torque nano oscillator in a linear sequence. The corresponding oscillation voltage is generated by the spin torque of the spin torque nano oscillator.

[0010] The nonlinear module is used to obtain the nonlinear part of the oscillating voltage amplitude, which has the same dimension as the linear part.

[0011] The splicing output module is used to splice the linear and nonlinear parts to obtain a complete feature vector. The complete feature vector is then multiplied by the output weights of the reservoir computing system based on spin torque nano-oscillators to obtain the target value corresponding to the external data. The output weights are obtained by linear regression training using prior external data through the reservoir computing system based on spin torque nano-oscillators.

[0012] In one embodiment, the above-described reservoir computing system based on a spin torque nano-oscillator further includes:

[0013] The current scaling module is used to scale the input current converted from the linear portion to a current range within which the spin torque nano oscillator can generate a stable oscillation voltage.

[0014] In one embodiment, during the linear regression training process using prior external data, the sampling time step of the output oscillation voltage amplitude of the reservoir computing system based on spin torque nanooscillators is kept consistent with the time interval of current injection.

[0015] On the other hand, a method for calculating a reservoir based on a spin torque nano-oscillator is also provided, including the following steps:

[0016] Based on the dimensions and time sequence of the external data to be processed, the external data is expanded to form the linear part of a one-dimensional feature vector;

[0017] After the linear portion is converted into an input current, it is injected into the spin torque nano oscillator in a linear sequence. The corresponding oscillation voltage is generated through the spin-transfer torque effect of the spin torque nano oscillator.

[0018] The amplitude of the oscillating voltage is obtained as a nonlinear component of the same dimension as the linear component;

[0019] The linear and nonlinear parts are concatenated to obtain a complete feature vector. The complete feature vector is then multiplied by the output weights of the reservoir computing system based on spin torque nano-oscillators to obtain the target value corresponding to the external data. The output weights are obtained by linear regression training using prior external data through the reservoir computing system based on spin torque nano-oscillators.

[0020] In one embodiment, before the linear portion is converted into an input current and injected into the spin torque nano-oscillator in a linear sequence, the above-mentioned reservoir calculation method further includes the following steps:

[0021] The input current converted from the linear portion is scaled to the current range within which the spin torque nano-oscillator can generate a stable oscillation voltage.

[0022] In another aspect, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described reservoir calculation method based on a spin torque nano-oscillator.

[0023] One of the above technical solutions has the following advantages and beneficial effects:

[0024] The aforementioned reservoir computing system, method, and medium based on spin-torque nano-oscillators (STNOs) directly generate nonlinear signals from linear signals by utilizing the inherent nonlinear current-voltage characteristics of STNOs, thus realizing the generation of the nonlinear part of eigenvectors in next-generation reservoir computing (NG-RC). By replacing traditional polynomial algorithms with physical laws, this scheme effectively solves three major problems in NG-RC: feature dimension combinatorial explosion, high computational overhead due to numerous matrix operations, and insufficient multi-scale dynamics capture capability. Compared with existing technologies, this scheme achieves adaptive frequency domain response at the hardware level (strong current triggers high-frequency oscillations to capture fast-changing processes, while weak current maintains low-frequency oscillations to focus on slow-changing dynamics), without the need for complex reconfiguration. Combined with the low computational overhead of analog domain processing (reusing linear dimensions and avoiding matrix operations), it can improve time-series processing efficiency without significantly increasing hardware costs, providing a new approach for building high-performance, high-energy-efficiency neuromorphic computing accelerators, suitable for applications such as speech recognition and time series analysis. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram of the module architecture of a reservoir computing system based on a spin torque nano-oscillator in one embodiment.

[0027] Figure 2 This is a schematic diagram of a reservoir computing scheme based on a spin torque nano-oscillator, consisting of linear reservoir nodes and a nonlinear output layer, in one embodiment.

[0028] Figure 3 This is a schematic diagram of an experimental setup for processing data using a spin torque nano-oscillator in one embodiment.

[0029] Figure 4 This is a schematic diagram of the offline training process of a reservoir computing experimental device based on a spin torque nano-oscillator, consisting of linear reservoir nodes and a nonlinear output layer, in one embodiment.

[0030] Figure 5 This is a flowchart illustrating a reservoir calculation method based on a spin torque nano-oscillator in one embodiment. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the 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 merely illustrative and not intended to limit the invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention.

[0032] It should be noted that, in this document, the reference to "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The presentation of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will understand that the embodiments described herein can be combined with other embodiments. The term "and / or" as used herein refers to any combination of one or more of the associated listed items, and all possible combinations, including such combinations.

[0033] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0034] In NG-RC, the linear part (Olin) of the feature vector is composed of observation data from the current and historical time steps, forming a time-delayed embedding sequence. The non-linear part (Ononlin) of the feature vector is a non-linear function of Olin. Existing research uses polynomial combinations of the linear feature vector Olin to generate higher-order non-linear terms. Therefore, the dimension of the non-linear feature vector (i.e., the dimension Dnonlin of the non-linear part) is determined by the polynomial order, the time series dimension of the original input data, and the number of time delay steps. When any parameter increases, the feature dimension grows geometrically, causing a feature dimension explosion problem. Furthermore, polynomials require a large number of matrix operations, resulting in high computational overhead and energy consumption. Therefore, the advantage of polynomial NG-RC is significant in low-dimensional systems, but its computational complexity increases non-linearly with dimension. Here, Dlin represents the dimension of the linear part.

[0035] However, complex systems require the simultaneous capture of dynamics at different time scales, but the polynomial order in polynomial methods must be pre-set manually and cannot be dynamically adjusted. For example, the Lorenz63 system (a simplified system describing atmospheric convection, exhibiting chaotic characteristics, and a classic model in nonlinear dynamical system research) uses a fixed-order second-order polynomial, while the double-vortex system, due to the presence of non-polynomial vector fields, raises it to third order. Fast-changing processes require high-frequency components, but the cutoff frequency of low-order polynomials is insufficient, leading to aliasing errors; slow-changing processes need to retain DC / low-frequency components, but higher-order terms are overly sensitive to low-frequency signals, amplifying the basis noise. While higher-order terms can extend the bandwidth, they can cause dimensionality explosion. Complex systems need to analyze both fast and slow-changing processes simultaneously, which static polynomials cannot adequately address.

[0036] In one embodiment, such as Figure 1 As shown, a reservoir computing system 100 based on a spin-torque nano-oscillator is provided. The spin-torque nano-oscillator serves as a nonlinear node in next-generation reservoir computing, used to generate the nonlinear part of the eigenvector in the next-generation reservoir computing. Figure 1 As shown, the reservoir computing system 100 may include a data expansion module 11, an injection oscillation module 13, a nonlinear module 15, and a splicing output module 17. The data expansion module 11 expands the external data to form a linear part of a one-dimensional feature vector based on the dimension and temporal relationship of the external data to be processed. The injection oscillation module 13 converts the linear part into an input current and injects it into a spin torque nano-oscillator in a linear sequence, generating a corresponding oscillation voltage through the spin-transfer torque of the spin torque nano-oscillator. The nonlinear module 15 obtains the amplitude of the oscillation voltage to form a nonlinear part with the same dimension as the linear part. The splicing output module 17 splices the linear and nonlinear parts to obtain a complete feature vector, and multiplies the complete feature vector by the output weights of the reservoir computing system based on the spin torque nano-oscillator to obtain the target value corresponding to the external data. The output weights are obtained by linear regression training of the reservoir computing system based on the spin torque nano-oscillator using prior external data.

[0037] It is understood that in this embodiment, the inherent nonlinear current-voltage characteristics of the spin torque nano-oscillator (STNO) are used to directly generate a nonlinear signal from a linear signal, thereby realizing the function of generating the nonlinear part of the eigenvector through polynomials in next-generation reservoir calculations. This design in this embodiment belongs to analog domain current-voltage conversion, using physical laws to replace algorithm design, thus resulting in extremely low computational overhead. In this embodiment, the spin torque nano-oscillator (STNO) in... Figure 1As a nonlinear node in the reservoir calculation, it utilizes its physical-level nonlinear dynamic characteristics (nonlinear current-voltage conversion) to introduce rich nonlinear expression capabilities to the reservoir, thereby enhancing the system's performance in processing complex time series (such as chaotic signals and communication signals). Compared with the nonlinearity of traditional software simulation, this hardware-level nonlinearity has advantages such as low power consumption, high speed and strong physical robustness.

[0038] It should be noted that, Figure 2 This is a schematic diagram of a reservoir computing scheme based on a spin torque nano-oscillator, consisting of linear reservoir nodes and a nonlinear output layer. Figure 2 The leftmost part represents the three-dimensional time series signal. x , y , z At a certain point in time , The data is sampled at each moment and used as the input sequence for the system. Figure 2 The middle section describes the internal process of the module (NG-RC with STNO): the first step is linear input generation. That is, first construct a linear input matrix containing the current time step. and the previous moment signal value This is the basic input for linear operations; then, a nonlinear transformation of the STNO is performed. The spin torque nano-oscillator (STNO) works as the core nonlinear element. First, a current I is input to the STNO. Utilizing its spintronic characteristics (such as the spin-transfer torque effect), the STNO converts the current into a voltage, completing the nonlinear mapping and generating a nonlinear output matrix. Corresponding to the transformed Finally, the total output and the prediction for the next time step are used to linearly output the results. and nonlinear output The combination (i.e., Ototal,i) is then processed using the output weights Wout to finally generate the next time step. Output It completes the processing of time series data (such as prediction and feature extraction). Figure 2 The rightmost section displays the system output, showing the time series signal after system processing. exist Output sequence at time step This demonstrates the system's ability to predict or transform the time series data for the next step. The magnetic moment of the free layer of the spin torque nano-oscillator STNO. The magnetic moment of the fixed layer of the spin torque nano oscillator STNO.

[0039] First, external data is expanded according to its own dimensions and temporal sequence to form the linear part of a one-dimensional feature vector. Then, the linear part is converted into current and injected into a spin torque nano-oscillator (STNO) in a linear sequence, generating a corresponding oscillation voltage through the spin-transfer torque (STT). Next, the amplitude of the oscillation voltage is obtained to form the nonlinear part, which inherits the dimension of the linear part, and both have the same dimension (i.e., Dlin = Dnonlin). Finally, the linear and nonlinear parts are concatenated to obtain the complete feature vector, whose dimension is the sum of the dimensions of the linear and nonlinear parts (i.e., Dtotal = Dlin + Dnonlin). A simple linear regression training is then performed to obtain the output weight matrix. The target value (i.e., the output sequence or signal corresponding to the external data) is obtained by multiplying the feature vector by the output weight.

[0040] The nonlinear part of the feature vector directly inherits the dimension of the linear part. By reusing the linear dimension architecture, the complexity of the overall feature space is significantly reduced. Furthermore, thanks to the inherent intrinsic relaxation characteristics of the spin torque nano oscillator STNO, the current state in the time delay sequence will dynamically integrate the residual effects of historical time data to form an adaptive time-series correlation mechanism.

[0041] The aforementioned voltage oscillation amplitude is robust to noise, which is attributed to its dual-torque dynamic balance mechanism: the driving effect of the injected current generates a positive torque that induces precession, while the dissipation effect of magnetic damping forms a reverse torque that hinders motion. When the intensity of the injected current exceeds a critical threshold, the two torques dynamically cancel each other out, forming a steady state similar to an energy trap. This antagonistic torque system automatically absorbs environmental noise disturbances (such as thermal fluctuations or electromagnetic interference), causing the output oscillation to lock onto a stable trajectory.

[0042] In dynamical system modeling, noise immunity enhancement refers to improving the model's robustness to input noise through specific mechanisms, ensuring that prediction results are not affected by weak perturbations. Traditional methods rely on Tikhonov regularization (also known as Ridge Regression) to suppress noise, but the spin torque nano-oscillator (STNO) achieves automatic regularization through hardware physical characteristics. Weak noise (such as sensor noise or environmental disturbances) usually manifests as low-amplitude currents that cannot trigger STNO oscillations. Therefore, these noise signals are naturally suppressed, and only effective signals are retained for nonlinear feature generation. In other words, low-amplitude noise is directly filtered out during the feature generation stage, which is equivalent to pre-filtering. This improves noise robustness, simplifies the noise processing process, and makes the optimization of the regularization strength α more stable (i.e., α is less sensitive to noise). The STNO reduces the difficulty of tuning α through hardware mechanisms.

[0043] The spin torque nano-oscillator (STNO) achieves hardware-level intelligent decoupling of multi-scale dynamics through its unique current-frequency coupling mechanism. Its core lies in the direct conversion of the input current intensity into a real-time control signal for the oscillation frequency: when a strong current is injected, the STNO is excited to generate high-frequency oscillations, and its response bandwidth automatically expands, thus accurately capturing fast-changing processes in the system; conversely, when the input current is weak, the STNO maintains low-frequency oscillations, and its response naturally focuses on the slow-changing dynamics of the system. This dynamic and continuous frequency-domain adaptive capability allows a single STNO to seamlessly cover multiple orders of magnitude of frequency range without any external reconfiguration or parameter adjustment. Essentially, it directly maps mathematical relationships to simulate physical processes, thus natively solving the modeling challenge of multi-scale coupling at the hardware level. It not only decouples fast and slow processes in real time but also automatically balances the transient dynamics and steady-state behavior of the system through its inherent physical response characteristics, providing a flexible and efficient solution for accurate modeling of complex systems.

[0044] The aforementioned reservoir computing system 100 based on a spin torque nano-oscillator (STNO) utilizes the inherent nonlinear current-voltage characteristics of STNOs to directly generate nonlinear signals from linear signals. This enables the generation of the nonlinear portion of feature vectors in next-generation reservoir computing, thereby replacing traditional polynomial algorithms with physical laws. This effectively solves three major problems in next-generation reservoir computing: feature dimension explosion, high computational overhead due to numerous matrix operations, and insufficient multi-scale dynamics capture capability. Compared to existing technologies, this solution achieves adaptive frequency domain response at the hardware level (strong current triggers high-frequency oscillations to capture fast-changing processes, while weak current maintains low-frequency oscillations to focus on slow-changing dynamics), eliminating the need for complex reconfiguration. Combined with the low computational overhead of analog domain processing (reusing linear dimensions and avoiding matrix operations), it can improve time-series processing efficiency without significantly increasing hardware costs. This provides a new approach for building high-performance, high-energy-efficiency neuromorphic computing accelerators, suitable for applications such as speech recognition and time series analysis.

[0045] Figure 3 This paper introduces the setup for experiments using a spin torque nano-oscillator (STNO). When processing external data using an STNO, a bias current must first be input. I dc The free-layer magnetization of the spin torque nano-oscillator (STNO) is itself subject to magnetic damping, much like frictional resistance, which causes any oscillation or precession to rapidly decay and stop, only when the bias current exceeds a specific threshold current. I thAt this point, the energy provided by the spin-transfer torque STT can completely offset the energy consumed by the damping, and magnetization can begin and maintain stable periodic precession (i.e., oscillation). The oscillation amplitude is strongly dependent on the bias current. Choosing an appropriate current value can set the oscillator to an optimal operating point with strong nonlinear response and high signal-to-noise ratio. This nonlinearity is key to simulating neuronal behavior and realizing complex calculations. Without a nonlinear oscillator, only the input signal (i.e., external data) can be linearly amplified, and the required complex transformations cannot be completed.

[0046] After external data is input, it is converted into a physical voltage waveform (Vin) by an arbitrary waveform generator (AWG). Then, the voltage is converted into an alternating current Iac by a voltage-controlled current source (VCCS). The DC current Idc and the AC current Iac pass through the inductor L and capacitor C in the biaser (Bias-Tee, also called a T-type biaser) to achieve the functions of blocking AC and passing DC, and blocking DC and passing AC respectively. Finally, they are superimposed and injected into the magnetic tunnel junction (MTJ). The amplitude of the oscillation voltage is measured by a microwave diode D0. Finally, the output data is sampled by an ADC sampling circuit for subsequent processing.

[0047] like Figure 4 As shown, the output weights are obtained through a separate offline training process. This offline training process first preprocesses the external data used as prior data to extract features more effective for the current task from the original data. The preprocessed external data is then used in subsequent operations to ensure the final effect. The preprocessing methods for different task data can vary. For example, for speech data, silence segments need to be removed, Fourier transforms are performed using frame-by-frame windowing to achieve time-frequency conversion, and MFCC (Mel-frequency cepstral coefficients) feature extraction is performed. For prediction tasks, lagging features are created, using data from the previous period (e.g., the previous day or several hours) as new features. Specifically, taking weather prediction as an example, temperature data from the previous three days (i.e., external data) is typically used as the feature for predicting the temperature Ti of the current day. These temperature data are uniformly expanded according to their dimension and temporal relationship, arranged into a one-dimensional linear vector. ,in, This indicates the temperature data for the 3rd day prior to the current day. This indicates the temperature data for the day before yesterday. This indicates the temperature data for the day preceding the current date.

[0048] In one embodiment, the above-described reservoir computing system 100 based on a spin torque nano oscillator may further include a current scaling module for scaling the input current converted from the linear portion to a current range within which the spin torque nano oscillator can generate a stable oscillating current.

[0049] It is understandable that different spin torque nano oscillators (STNOs) generate stable oscillating currents within different ranges. Therefore, external data needs to be scaled to the corresponding current range according to the specific STNO requirements. The scaled current is then ordered sequentially. Injected into a spin torque nano-oscillator (STNO), wherein, express The corresponding input current. express The corresponding input current, express The corresponding input current. The time interval for data point injection also needs to be adjusted according to the device characteristics of the spin torque nano-oscillator (STNO): for example, the magnetization precession of the free layer of the STNO has an intrinsic relaxation time. If the injection time interval is too short, the magnetization state will not be stable, which will cause the oscillation amplitude to not reach a steady state, resulting in nonlinear response distortion and a decrease in noise suppression capability. If the injection time interval is too long, the correlation between sequence points will decay, resulting in insufficient equivalent memory depth of the reservoir.

[0050] Furthermore, the sampling time step of the output oscillation voltage amplitude is consistent with the time interval of the current injection. In this reservoir computing system based on a spin torque nano-oscillator, a linear input corresponds to a nonlinear output, i.e., the current... Current and current Corresponding to the output voltage amplitude Output voltage amplitude and output voltage amplitude After concatenating the input current and output voltage amplitudes, the final feature vector (Ii-3, Ii-2, Ii-1, Vi-3, Vi-2, Vi-1) can be formed. Next, training and testing datasets can be constructed separately (e.g., the datasets can be divided in a 7:3 or 6:4 ratio). Different training methods are used depending on the specific task requirements (e.g., classification, time series prediction, or regression analysis). For example, for weather prediction, a time series prediction method is used to predict the temperature of the current day using temperature data from the previous three days. The predicted temperature value and the actual temperature of the current day are then compared using ridge regression (a biased estimation regression method used to solve the problem of collinear data) to calculate the output weight Wout. Subsequently, the output weight Wout is iteratively updated using all prior data in the training dataset until the end of the sequence. Finally, the prediction accuracy of this reservoir computing system based on a spin torque nano-oscillator is tested using the testing dataset to determine whether its prediction accuracy meets the design requirements. If it meets the design requirements, it can be used to process real-time time series analysis tasks.

[0051] Each module in the aforementioned reservoir computing system 100 based on spin torque nano-oscillators can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in hardware or independently of a device with data processing capabilities, or stored in software within the memory of the aforementioned device, so that the processor can call and execute the operations corresponding to each module. The aforementioned device can be, but is not limited to, various types of data processing devices already existing in the art.

[0052] In one embodiment, such as Figure 5 As shown, a method for calculating a reservoir based on a spin torque nano-oscillator is also provided, which may include the following steps S12 to S18:

[0053] S12, based on the dimension and time sequence relationship of the external data to be processed, expand the external data to form the linear part of a one-dimensional feature vector;

[0054] S14 converts the linear portion into an input current and injects it into the spin torque nano oscillator in a linear sequence. The corresponding oscillation voltage is generated through the spin-transfer torque effect of the spin torque nano oscillator.

[0055] S16, obtain the nonlinear part of the oscillation voltage amplitude with the same dimension as the linear part;

[0056] S18: The linear and nonlinear parts are concatenated to obtain a complete feature vector. The complete feature vector is multiplied by the output weights of the reservoir computing system based on spin torque nano-oscillators to obtain the target value corresponding to the external data. The output weights are obtained by linear regression training using prior external data through the reservoir computing system based on spin torque nano-oscillators.

[0057] The aforementioned reservoir computation method based on spin-torque nano-oscillators (STNOs) directly generates nonlinear signals from linear signals by utilizing the inherent nonlinear current-voltage characteristics of STNOs, thus realizing the generation of the nonlinear part of the eigenvectors in next-generation reservoir computation (NG-RC). By replacing traditional polynomial algorithms with physical laws, this scheme effectively solves three major problems in NG-RC: feature dimension combinatorial explosion, high computational overhead due to numerous matrix operations, and insufficient multi-scale dynamics capture capability. Compared with existing technologies, this scheme achieves adaptive frequency domain response at the hardware level (strong current triggers high-frequency oscillations to capture fast-changing processes, while weak current maintains low-frequency oscillations to focus on slow-changing dynamics), without the need for complex reconfiguration. Combined with the low computational overhead of analog domain processing (reusing linear dimensions and avoiding matrix operations), it can improve time-series processing efficiency without significantly increasing hardware costs, providing a new approach for building high-performance, high-energy-efficiency neuromorphic computing accelerators, suitable for applications such as speech recognition and time series analysis.

[0058] In one embodiment, before converting the linear portion into an input current and injecting it into the spin torque nano-oscillator in a linear sequence, the above-mentioned reservoir calculation method further includes the step of:

[0059] The input current converted from the linear portion is scaled to the current range within which the spin torque nano-oscillator can generate a stable oscillation voltage.

[0060] In one embodiment, during the linear regression training process using prior external data in the reservoir computing system based on spin torque nano-oscillators, the sampling time step of the output oscillation voltage amplitude is kept consistent with the time interval of current injection.

[0061] It is understood that the explanations of the features in the above-mentioned reservoir calculation method based on spin torque nano-oscillators can be understood by referring to the corresponding explanations in the various embodiments of the reservoir calculation system 100 based on spin torque nano-oscillators.

[0062] It should be understood that, although Figure 5 The steps are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order in which these steps are executed; they can be performed in other orders. Figure 5 At least some 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 sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0063] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, it performs the following processing steps: Based on the dimension and temporal relationship of the external data to be processed, the external data is expanded to form a linear part of a one-dimensional feature vector; the linear part is converted into an input current and injected into a spin torque nano-oscillator in a linear order, generating a corresponding oscillation voltage through the spin-transfer torque of the spin torque nano-oscillator; the amplitude of the oscillation voltage is obtained to form a nonlinear part of the same dimension as the linear part; the linear part and the nonlinear part are concatenated to obtain a complete feature vector; the complete feature vector is multiplied by the output weights of a reservoir computing system based on the spin torque nano-oscillator to obtain the target value corresponding to the external data; wherein, the output weights are obtained by linear regression training using prior external data through a reservoir computing system based on the spin torque nano-oscillator.

[0064] In one embodiment, when the computer program is executed by the processor, it can also implement the steps or sub-steps added to the various embodiments of the above-described reservoir calculation method based on spin torque nano-oscillators.

[0065] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this 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. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), memory bus DRAM (RDRAM), and interface DRAM (DRDRAM), etc.

[0066] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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.

[0067] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of protection of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and all such modifications and improvements fall within the scope of protection of the present invention.

Claims

1. A spin-torque nanowire oscillator based reservoir computing system, comprising: Spin-torque nano-oscillators as non-linear nodes of next-generation reservoir computing for generating non-linear part of feature vectors in next-generation reservoir computing; The reservoir computing system comprises: a data unfolding module configured to unfold external data to be processed according to a dimension and a time sequence relationship of the external data to form a linear part of a one-dimensional feature vector; an injection oscillation module configured to inject the linear part into the spin-torque nano-oscillator in a linear order after the linear part is converted into an input current, and generate a corresponding oscillation voltage through a spin-transfer torque effect of the spin-torque nano-oscillator; a non-linear module configured to obtain a non-linear part with the same dimension as the linear part by using amplitudes of the oscillation voltage; a splicing output module configured to splice the linear part and the non-linear part to obtain a complete feature vector, and multiply the complete feature vector by an output weight of the reservoir computing system based on the spin-torque nano-oscillator to obtain a target value corresponding to the external data; wherein the output weight is obtained by linear regression training of the reservoir computing system based on the spin-torque nano-oscillator using prior external data.

2. The spin-torque nanowire oscillator-based reservoir computing system of claim 1, wherein, Further comprising: a current scaling module configured to scale the input current converted from the linear part to a current range in which the spin-torque nano-oscillator can generate a stable oscillation voltage.

3. The spin-torque nanowire oscillator-based reservoir computing system of claim 1 or 2, wherein, In the process of linear regression training of the reservoir computing system based on the spin-torque nano-oscillator using prior external data, a sampling time step of amplitudes of the output oscillation voltage is consistent with a time interval of current injection.

4. A spin-torque nano-oscillator based reservoir computing method, characterized in that, The method comprises the steps of: unfolding external data to be processed according to a dimension and a time sequence relationship of the external data to form a linear part of a one-dimensional feature vector; injecting the linear part into the spin-torque nano-oscillator in a linear order after the linear part is converted into an input current, and generating a corresponding oscillation voltage through a spin-transfer torque effect of the spin-torque nano-oscillator; obtaining a non-linear part with the same dimension as the linear part by using amplitudes of the oscillation voltage; splicing the linear part and the non-linear part to obtain a complete feature vector, and multiplying the complete feature vector by an output weight of the reservoir computing system based on the spin-torque nano-oscillator to obtain a target value corresponding to the external data; wherein the output weight is obtained by linear regression training of the reservoir computing system based on the spin-torque nano-oscillator using prior external data.

5. The spin-torque nanowire oscillator-based reservoir computing method of claim 4, wherein, Before injecting the linear part into the spin-torque nano-oscillator in a linear order after the linear part is converted into an input current, the reservoir computing method further comprises the step of: scaling the input current converted from the linear part to a current range in which the spin-torque nano-oscillator can generate a stable oscillation voltage.

6. The spin-torque nanowire oscillator-based reservoir computing method according to claim 4 or 5, c h a r a c t e r i z e d b y, In the process of linear regression training of the reservoir computing system based on the spin-torque nano-oscillator using prior external data, a sampling time step of amplitudes of the output oscillation voltage is consistent with a time interval of current injection.

7. A computer readable storage medium having stored thereon a computer program, characterized in that The computer program, when executed by a processor, implements the steps of the reservoir computing method based on the spin-torque nano-oscillator according to any one of claims 4 to 6.

Images