Tuning Convolutional Spiking Neural Network With Biologically Plausible Reward Propagation

A Reinforced, Event-Driven, and Attention-Based Convolution Spiking Neural Network for Multivariate Time Series Prediction

Despite spiking neural networks (SNNs) inherently exceling at processing time series due to their rich spatio-temporal information and efficient event-driven computing, the challenge of extracting complex correlations between variables in multivariate time series (MTS) remains to be addressed. This paper proposes a reinforced, event-driven, and attention-based convolution SNN model (REAT-CSNN) with three novel features. First, a joint Gramian Angular Field and Rate (GAFR) coding scheme is proposed to convert MTS into spike images, preserving the inherent features in MTS, such as the temporal patterns and spatio-temporal correlations between time series. Second, an advanced LIF-pooling strategy is developed, which is then theoretically and empirically proved to be effective in preserving more features from the regions of interest in spike images than average-pooling strategies. Third, a convolutional block attention mechanism (CBAM) is redesigned to support spike-based input, enhancing event-driven characteristics in weighting operations while maintaining outstanding capability to capture the information encoded in spike images. Experiments on multiple MTS data sets, such as stocks and PM2.5 data sets, demonstrate that our model rivals, and even surpasses, some CNN- and RNN-based techniques, with up to 3% better performance, while consuming significantly less energy.

Efficient Deep Spiking Multilayer Perceptrons With Multiplication-Free Inference

Advancements in adapting deep convolution architectures for spiking neural networks (SNNs) have significantly enhanced image classification performance and reduced computational burdens. However, the inability of multiplication-free inference (MFI) to align with attention and transformer mechanisms, which are critical to superior performance on high-resolution vision tasks, imposes limitations on these gains. To address this, our research explores a new pathway, drawing inspiration from the progress made in multilayer perceptrons (MLPs). We propose an innovative spiking MLP architecture that uses batch normalization (BN) to retain MFI compatibility and introduce a spiking patch encoding (SPE) layer to enhance local feature extraction capabilities. As a result, we establish an efficient multistage spiking MLP network that blends effectively global receptive fields with local feature extraction for comprehensive spike-based computation. Without relying on pretraining or sophisticated SNN training techniques, our network secures a top-one accuracy of 66.39% on the ImageNet-1K dataset, surpassing the directly trained spiking ResNet-34 by 2.67%. Furthermore, we curtail computational costs, model parameters, and simulation steps. An expanded version of our network compares with the performance of the spiking VGG-16 network with a 71.64% top-one accuracy, all while operating with a model capacity 2.1 times smaller. Our findings highlight the potential of our deep SNN architecture in effectively integrating global and local learning abilities. Interestingly, the trained receptive field in our network mirrors the activity patterns of cortical cells.

A Gradient-Guided Evolutionary Neural Architecture Search

Neural architecture search (NAS) is a popular method that can automatically design deep neural network structures. However, designing a neural network using NAS is computationally expensive. This article proposes a gradient-guided evolutionary NAS (GENAS) to design convolutional neural networks (CNNs) for image classification. GENAS is a hybrid algorithm that combines evolutionary global and local search operators to evolve a population of subnets sampled from a supernet. Each candidate architecture is encoded as a table describing which operations are associated with the edges between nodes signifying feature maps. Besides, evolutionary optimization uses novel crossover and mutation operators to manipulate the subnets using the proposed tabular encoding. Every generations, the candidate architectures undergo a local search inspired by differentiable NAS. GENAS is designed to overcome the limitations of both evolutionary and gradient descent NAS. This algorithmic structure enables the performance assessment of the candidate architecture without retraining, thus limiting the NAS calculation time. Furthermore, subnet individuals are decoupled during evaluation to prevent strong coupling of operations in the supernet. The experimental results indicate that the searched structures achieve test errors of 2.45%, 16.86%, and 23.9% on CIFAR-10/100/ImageNet datasets and it costs only 0.26 GPU days on a graphic card. GENAS can effectively expedite the training and evaluation processes and obtain high-performance network structures.

Spiking Neural Network for Ultralow-Latency and High-Accurate Object Detection

Spiking Neural Networks (SNNs) have attracted significant attention for their energy-efficient and brain-inspired event-driven properties. Recent advancements, notably Spiking-YOLO, have enabled SNNs to undertake advanced object detection tasks. Nevertheless, these methods often suffer from increased latency and diminished detection accuracy, rendering them less suitable for latency-sensitive mobile platforms. Additionally, the conversion of artificial neural networks (ANNs) to SNNs frequently compromises the integrity of the ANNs' structure, resulting in poor feature representation and heightened conversion errors. To address the issues of high latency and low detection accuracy, we introduce two solutions: timestep compression and spike-time-dependent integrated (STDI) coding. Timestep compression effectively reduces the number of timesteps required in the ANN-to-SNN conversion by condensing information. The STDI coding employs a time-varying threshold to augment information capacity. Furthermore, we have developed an SNN-based spatial pyramid pooling (SPP) structure, optimized to preserve the network's structural efficacy during conversion. Utilizing these approaches, we present the ultralow latency and highly accurate object detection model, SUHD. SUHD exhibits exceptional performance on challenging datasets like PASCAL VOC and MS COCO, achieving a remarkable reduction of approximately 750 times in timesteps and a 30% enhancement in mean average precision (mAP) compared to Spiking-YOLO on MS COCO. To the best of our knowledge, SUHD is currently the deepest spike-based object detection model, achieving ultralow timesteps for lossless conversion.

A novel approach to enhancing biomedical signal recognition via hybrid high-order information bottleneck driven spiking neural networks

Biomedical signals, encapsulating vital physiological information, are pivotal in elucidating human traits and conditions, serving as a cornerstone for advancing human-machine interfaces. Nonetheless, the fidelity of biomedical signal interpretation is frequently compromised by pervasive noise sources such as skin, motion, and equipment interference, posing formidable challenges to precision recognition tasks. Concurrently, the burgeoning adoption of intelligent wearable devices illuminates a societal shift towards enhancing life and work through technological integration. This surge in popularity underscores the imperative for efficient, noise-resilient biomedical signal recognition methodologies, a quest that is both challenging and profoundly impactful. This study proposes a novel approach to enhancing biomedical signal recognition. The proposed approach employs a hierarchical information bottleneck mechanism within SNNs, quantifying the mutual information in different orders based on the depth of information flow in the network. Subsequently, these mutual information, together with the network's output and category labels, are restructured based on information theory principles to form the loss function used for training. A series of theoretical analyses and substantial experimental results have shown that this method can effectively compress noise in the data, and on the premise of low computational cost, it can also significantly outperform its vanilla counterpart in terms of classification performance.

Toward High-Accuracy and Low-Latency Spiking Neural Networks With Two-Stage Optimization

Spiking neural networks (SNNs) operating with asynchronous discrete events show higher energy efficiency with sparse computation. A popular approach for implementing deep SNNs is artificial neural network (ANN)-SNN conversion combining both efficient training of ANNs and efficient inference of SNNs. However, the accuracy loss is usually nonnegligible, especially under few time steps, which restricts the applications of SNN on latency-sensitive edge devices greatly. In this article, we first identify that such performance degradation stems from the misrepresentation of the negative or overflow residual membrane potential in SNNs. Inspired by this, we decompose the conversion error into three parts: quantization error, clipping error, and residual membrane potential representation error. With such insights, we propose a two-stage conversion algorithm to minimize those errors, respectively. In addition, we show that each stage achieves significant performance gains in a complementary manner. By evaluating on challenging datasets including CIFAR- 10, CIFAR- 100, and ImageNet, the proposed method demonstrates the state-of-the-art performance in terms of accuracy, latency, and energy preservation. Furthermore, our method is evaluated using a more challenging object detection task, revealing notable gains in regression performance under ultralow latency, when compared with existing spike-based detection algorithms. Codes will be available at: https://github.com/Windere/snn-cvt-dual-phase.

Defect-Tolerant Memristor Crossbar Circuits for Local Learning Neural Networks

Local learning algorithms, such as Equilibrium Propagation (EP), have emerged as alternatives to global learning methods like backpropagation for training neural networks. EP offers the potential for more energy-efficient hardware implementation by utilizing only local neuron information for weight updates. However, the practical implementation of EP using memristor-based circuits has significant challenges due to the immature fabrication processes of memristors, resulting in defects and variability issues. Previous implementations of EP with memristor crossbars use two separate circuits for the free and nudge phases. This approach can suffer differences in defects and variability between the two circuits, potentially leading to significant performance degradation. To overcome these limitations, in this paper, we propose a novel time-multiplexing technique that combines the free and nudge phases into a single memristor circuit. Our proposed scheme integrates the dynamic equations of the free and nudge phases into one circuit, allowing defects and variability compensation during the training. Simulations using the MNIST dataset demonstrate that our approach maintains a 92% recognition rate even with a 10% defect rate in memristors, compared to 33% for the previous scheme. Furthermore, the proposed circuit reduces area overhead for both the memristor circuit solving EP's algorithm and the weight-update control circuit.

Event-Driven Spiking Learning Algorithm Using Aggregated Labels

Traditional spiking learning algorithm aims to train neurons to spike at a specific time or on a particular frequency, which requires precise time and frequency labels in the training process. While in reality, usually only aggregated labels of sequential patterns are provided. The aggregate-label (AL) learning is proposed to discover these predictive features in distracting background streams only by aggregated spikes. It has achieved much success recently, but it is still computationally intensive and has limited use in deep networks. To address these issues, we propose an event-driven spiking aggregate learning algorithm (SALA) in this article. Specifically, to reduce the computational complexity, we improve the conventional spike-threshold-surface (STS) calculation in AL learning by analytical calculating voltage peak values in spiking neurons. Then we derive the algorithm to multilayers by event-driven strategy using aggregated spikes. We conduct comprehensive experiments on various tasks including temporal clue recognition, segmented and continuous speech recognition, and neuromorphic image classification. The experimental results demonstrate that the new STS method improves the efficiency of AL learning significantly, and the proposed algorithm outperforms the conventional spiking algorithm in various temporal clue recognition tasks.

Self-Lateral Propagation Elevates Synaptic Modifications in Spiking Neural Networks for the Efficient Spatial and Temporal Classification

The brain's mystery for efficient and intelligent computation hides in the neuronal encoding, functional circuits, and plasticity principles in natural neural networks. However, many plasticity principles have not been fully incorporated into artificial or spiking neural networks (SNNs). Here, we report that incorporating a novel feature of synaptic plasticity found in natural networks, whereby synaptic modifications self-propagate to nearby synapses, named self-lateral propagation (SLP), could further improve the accuracy of SNNs in three benchmark spatial and temporal classification tasks. The SLP contains lateral pre ( SLP pre ) and lateral post ( SLP post ) synaptic propagation, describing the spread of synaptic modifications among output synapses made by axon collaterals or among converging synapses on the postsynaptic neuron, respectively. The SLP is biologically plausible and can lead to a coordinated synaptic modification within layers that endow higher efficiency without losing much accuracy. Furthermore, the experimental results showed the impressive role of SLP in sharpening the normal distribution of synaptic weights and broadening the more uniform distribution of misclassified samples, which are both considered essential for understanding the learning convergence and network generalization of neural networks.

Memristive leaky integrate-and-fire neuron and learnable straight-through estimator in spiking neural networks

Compared to artificial neural networks (ANNs), spiking neural networks (SNNs) present a more biologically plausible model of neural system dynamics. They rely on sparse binary spikes to communicate information and operate in an asynchronous, event-driven manner. Despite the high heterogeneity of the neural system at the neuronal level, most current SNNs employ the widely used leaky integrate-and-fire (LIF) neuron model, which assumes uniform membrane-related parameters throughout the entire network. This approach hampers the expressiveness of spiking neurons and restricts the diversity of neural dynamics. In this paper, we propose replacing the resistor in the LIF model with a discrete memristor to obtain the heterogeneous memristive LIF (MLIF) model. The memristance of the discrete memristor is determined by the voltage and flux at its terminals, leading to dynamic changes in the membrane time parameter of the MLIF model. SNNs composed of MLIF neurons can not only learn synaptic weights but also adaptively change membrane time parameters according to the membrane potential of the neuron, enhancing the learning ability and expression of SNNs. Furthermore, since the proper threshold of spiking neurons can improve the information capacity of SNNs, a learnable straight-through estimator (LSTE) is proposed. The LSTE, based on the straight-through estimator (STE) surrogate function, features a learnable threshold that facilitates the backward propagation of gradients through neurons firing spikes. Extensive experiments on several popular static and neuromorphic benchmark datasets demonstrate the effectiveness of the proposed MLIF and LSTE, especially on the DVS-CIFAR10 dataset, where we achieved the top-1 accuracy of 84.40 % .

Paired competing neurons improving STDP supervised local learning in Spiking Neural Networks

Direct training of Spiking Neural Networks (SNNs) on neuromorphic hardware has the potential to significantly reduce the energy consumption of artificial neural network training. SNNs trained with Spike Timing-Dependent Plasticity (STDP) benefit from gradient-free and unsupervised local learning, which can be easily implemented on ultra-low-power neuromorphic hardware. However, classification tasks cannot be performed solely with unsupervised STDP. In this paper, we propose Stabilized Supervised STDP (S2-STDP), a supervised STDP learning rule to train the classification layer of an SNN equipped with unsupervised STDP for feature extraction. S2-STDP integrates error-modulated weight updates that align neuron spikes with desired timestamps derived from the average firing time within the layer. Then, we introduce a training architecture called Paired Competing Neurons (PCN) to further enhance the learning capabilities of our classification layer trained with S2-STDP. PCN associates each class with paired neurons and encourages neuron specialization toward target or non-target samples through intra-class competition. We evaluate our methods on image recognition datasets, including MNIST, Fashion-MNIST, and CIFAR-10. Results show that our methods outperform state-of-the-art supervised STDP learning rules, for comparable architectures and numbers of neurons. Further analysis demonstrates that the use of PCN enhances the performance of S2-STDP, regardless of the hyperparameter set and without introducing any additional hyperparameters.

Application of Event Cameras and Neuromorphic Computing to VSLAM: A Survey

Simultaneous Localization and Mapping (SLAM) is a crucial function for most autonomous systems, allowing them to both navigate through and create maps of unfamiliar surroundings. Traditional Visual SLAM, also commonly known as VSLAM, relies on frame-based cameras and structured processing pipelines, which face challenges in dynamic or low-light environments. However, recent advancements in event camera technology and neuromorphic processing offer promising opportunities to overcome these limitations. Event cameras inspired by biological vision systems capture the scenes asynchronously, consuming minimal power but with higher temporal resolution. Neuromorphic processors, which are designed to mimic the parallel processing capabilities of the human brain, offer efficient computation for real-time data processing of event-based data streams. This paper provides a comprehensive overview of recent research efforts in integrating event cameras and neuromorphic processors into VSLAM systems. It discusses the principles behind event cameras and neuromorphic processors, highlighting their advantages over traditional sensing and processing methods. Furthermore, an in-depth survey was conducted on state-of-the-art approaches in event-based SLAM, including feature extraction, motion estimation, and map reconstruction techniques. Additionally, the integration of event cameras with neuromorphic processors, focusing on their synergistic benefits in terms of energy efficiency, robustness, and real-time performance, was explored. The paper also discusses the challenges and open research questions in this emerging field, such as sensor calibration, data fusion, and algorithmic development. Finally, the potential applications and future directions for event-based SLAM systems are outlined, ranging from robotics and autonomous vehicles to augmented reality.

Brain topology improved spiking neural network for efficient reinforcement learning of continuous control

The brain topology highly reflects the complex cognitive functions of the biological brain after million-years of evolution. Learning from these biological topologies is a smarter and easier way to achieve brain-like intelligence with features of efficiency, robustness, and flexibility. Here we proposed a brain topology-improved spiking neural network (BT-SNN) for efficient reinforcement learning. First, hundreds of biological topologies are generated and selected as subsets of the Allen mouse brain topology with the help of the Tanimoto hierarchical clustering algorithm, which has been widely used in analyzing key features of the brain connectome. Second, a few biological constraints are used to filter out three key topology candidates, including but not limited to the proportion of node functions (e.g., sensation, memory, and motor types) and network sparsity. Third, the network topology is integrated with the hybrid numerical solver-improved leaky-integrated and fire neurons. Fourth, the algorithm is then tuned with an evolutionary algorithm named adaptive random search instead of backpropagation to guide synaptic modifications without affecting raw key features of the topology. Fifth, under the test of four animal-survival-like RL tasks (i.e., dynamic controlling in Mujoco), the BT-SNN can achieve higher scores than not only counterpart SNN using random topology but also some classical ANNs (i.e., long-short-term memory and multi-layer perception). This result indicates that the research effort of incorporating biological topology and evolutionary learning rules has much in store for the future.

Low-variance Forward Gradients using Direct Feedback Alignment and momentum

Supervised learning in deep neural networks is commonly performed using error backpropagation. However, the sequential propagation of errors during the backward pass limits its scalability and applicability to low-powered neuromorphic hardware. Therefore, there is growing interest in finding local alternatives to backpropagation. Recently proposed methods based on forward-mode automatic differentiation suffer from high variance in large deep neural networks, which affects convergence. In this paper, we propose the Forward Direct Feedback Alignment algorithm that combines Activity-Perturbed Forward Gradients with Direct Feedback Alignment and momentum. We provide both theoretical proofs and empirical evidence that our proposed method achieves lower variance than forward gradient techniques. In this way, our approach enables faster convergence and better performance when compared to other local alternatives to backpropagation and opens a new perspective for the development of online learning algorithms compatible with neuromorphic systems.

A brain-inspired algorithm that mitigates catastrophic forgetting of artificial and spiking neural networks with low computational cost

Neuromodulators in the brain act globally at many forms of synaptic plasticity, represented as metaplasticity, which is rarely considered by existing spiking (SNNs) and nonspiking artificial neural networks (ANNs). Here, we report an efficient brain-inspired computing algorithm for SNNs and ANNs, referred to here as neuromodulation-assisted credit assignment (NACA), which uses expectation signals to induce defined levels of neuromodulators to selective synapses, whereby the long-term synaptic potentiation and depression are modified in a nonlinear manner depending on the neuromodulator level. The NACA algorithm achieved high recognition accuracy with substantially reduced computational cost in learning spatial and temporal classification tasks. Notably, NACA was also verified as efficient for learning five different class continuous learning tasks with varying degrees of complexity, exhibiting a markedly mitigated catastrophic forgetting at low computational cost. Mapping synaptic weight changes showed that these benefits could be explained by the sparse and targeted synaptic modifications attributed to expectation-based global neuromodulation.

Incorporating structural plasticity into self-organization recurrent networks for sequence learning

Introduction

Spiking neural networks (SNNs), inspired by biological neural networks, have received a surge of interest due to its temporal encoding. Biological neural networks are driven by multiple plasticities, including spike timing-dependent plasticity (STDP), structural plasticity, and homeostatic plasticity, making network connection patterns and weights to change continuously during the lifecycle. However, it is unclear how these plasticities interact to shape neural networks and affect neural signal processing.

Method

Here, we propose a reward-modulated self-organization recurrent network with structural plasticity (RSRN-SP) to investigate this issue. Specifically, RSRN-SP uses spikes to encode information, and incorporate multiple plasticities including reward-modulated spike timing-dependent plasticity (R-STDP), homeostatic plasticity, and structural plasticity. On the one hand, combined with homeostatic plasticity, R-STDP is presented to guide the updating of synaptic weights. On the other hand, structural plasticity is utilized to simulate the growth and pruning of synaptic connections.

Results and discussion

Extensive experiments for sequential learning tasks are conducted to demonstrate the representational ability of the RSRN-SP, including counting task, motion prediction, and motion generation. Furthermore, the simulations also indicate that the characteristics arose from the RSRN-SP are consistent with biological observations.

Memristor Crossbar Circuits Implementing Equilibrium Propagation for On-Device Learning

Equilibrium propagation (EP) has been proposed recently as a new neural network training algorithm based on a local learning concept, where only local information is used to calculate the weight update of the neural network. Despite the advantages of local learning, numerical iteration for solving the EP dynamic equations makes the EP algorithm less practical for realizing edge intelligence hardware. Some analog circuits have been suggested to solve the EP dynamic equations physically, not numerically, using the original EP algorithm. However, there are still a few problems in terms of circuit implementation: for example, the need for storing the free-phase solution and the lack of essential peripheral circuits for calculating and updating synaptic weights. Therefore, in this paper, a new analog circuit technique is proposed to realize the EP algorithm in practical and implementable hardware. This work has two major contributions in achieving this objective. First, the free-phase and nudge-phase solutions are calculated by the proposed analog circuits simultaneously, not at different times. With this process, analog voltage memories or digital memories with converting circuits between digital and analog domains for storing the free-phase solution temporarily can be eliminated in the proposed EP circuit. Second, a simple EP learning rule relying on a fixed amount of conductance change per programming pulse is newly proposed and implemented in peripheral circuits. The modified EP learning rule can make the weight update circuit practical and implementable without requiring the use of a complicated program verification scheme. The proposed memristor conductance update circuit is simulated and verified for training synaptic weights on memristor crossbars. The simulation results showed that the proposed EP circuit could be used for realizing on-device learning in edge intelligence hardware.

Explaining cocktail party effect and McGurk effect with a spiking neural network improved by Motif-topology

Network architectures and learning principles have been critical in developing complex cognitive capabilities in artificial neural networks (ANNs). Spiking neural networks (SNNs) are a subset of ANNs that incorporate additional biological features such as dynamic spiking neurons, biologically specified architectures, and efficient and useful paradigms. Here we focus more on network architectures in SNNs, such as the meta operator called 3-node network motifs, which is borrowed from the biological network. We proposed a Motif-topology improved SNN (M-SNN), which is further verified efficient in explaining key cognitive phenomenon such as the cocktail party effect (a typical noise-robust speech-recognition task) and McGurk effect (a typical multi-sensory integration task). For M-SNN, the Motif topology is obtained by integrating the spatial and temporal motifs. These spatial and temporal motifs are first generated from the pre-training of spatial (e.g., MNIST) and temporal (e.g., TIDigits) datasets, respectively, and then applied to the previously introduced two cognitive effect tasks. The experimental results showed a lower computational cost and higher accuracy and a better explanation of some key phenomena of these two effects, such as new concept generation and anti-background noise. This mesoscale network motifs topology has much room for the future.