Synaptic Plasticity Dynamics for Deep Continuous Local Learning (DECOLLE)

Emergence of brain-inspired small-world spiking neural network through neuroevolution

Studies suggest that the brain's high efficiency and low energy consumption may be closely related to its small-world topology and critical dynamics. However, existing efforts on the performance-oriented structural evolution of spiking neural networks (SNNs) are time-consuming and ignore the core structural properties of the brain. Here, we introduce a multi-objective Evolutionary Liquid State Machine (ELSM), which blends the small-world coefficient and criticality to evolve models and guide the emergence of brain-inspired, efficient structures. Experiments reveal ELSM's consistent and comparable performance, achieving 97.23% on NMNIST and outperforming LSM models on MNIST and Fashion-MNIST with 98.12% and 88.81% accuracies, respectively. Further analysis shows its versatility and spontaneous evolution of topologies such as hub nodes, short paths, long-tailed degree distributions, and numerous communities. This study evolves recurrent spiking neural networks into brain-inspired energy-efficient structures, showcasing versatility in multiple tasks and potential for adaptive general artificial intelligence.

SHIP: a computational framework for simulating and validating novel technologies in hardware spiking neural networks

Investigations in the field of spiking neural networks (SNNs) encompass diverse, yet overlapping, scientific disciplines. Examples range from purely neuroscientific investigations, researches on computational aspects of neuroscience, or applicative-oriented studies aiming to improve SNNs performance or to develop artificial hardware counterparts. However, the simulation of SNNs is a complex task that can not be adequately addressed with a single platform applicable to all scenarios. The optimization of a simulation environment to meet specific metrics often entails compromises in other aspects. This computational challenge has led to an apparent dichotomy of approaches, with model-driven algorithms dedicated to the detailed simulation of biological networks, and data-driven algorithms designed for efficient processing of large input datasets. Nevertheless, material scientists, device physicists, and neuromorphic engineers who develop new technologies for spiking neuromorphic hardware solutions would find benefit in a simulation environment that borrows aspects from both approaches, thus facilitating modeling, analysis, and training of prospective SNN systems. This manuscript explores the numerical challenges deriving from the simulation of spiking neural networks, and introduces SHIP, Spiking (neural network) Hardware In PyTorch, a numerical tool that supports the investigation and/or validation of materials, devices, small circuit blocks within SNN architectures. SHIP facilitates the algorithmic definition of the models for the components of a network, the monitoring of states and output of the modeled systems, and the training of the synaptic weights of the network, by way of user-defined unsupervised learning rules or supervised training techniques derived from conventional machine learning. SHIP offers a valuable tool for researchers and developers in the field of hardware-based spiking neural networks, enabling efficient simulation and validation of novel technologies.

A spiking neural network with continuous local learning for robust online brain machine interface

Objective.Spiking neural networks (SNNs) are powerful tools that are well suited for brain machine interfaces (BMI) due to their similarity to biological neural systems and computational efficiency. They have shown comparable accuracy to state-of-the-art methods, but current training methods require large amounts of memory, and they cannot be trained on a continuous input stream without pausing periodically to perform backpropagation. An ideal BMI should be capable training continuously without interruption to minimize disruption to the user and adapt to changing neural environments.Approach.We propose a continuous SNN weight update algorithm that can be trained to perform regression learning with no need for storing past spiking events in memory. As a result, the amount of memory needed for training is constant regardless of the input duration. We evaluate the accuracy of the network on recordings of neural data taken from the premotor cortex of a primate performing reaching tasks. Additionally, we evaluate the SNN in a simulated closed loop environment and observe its ability to adapt to sudden changes in the input neural structure.Main results.The continuous learning SNN achieves the same peak correlation (ρ=0.7) as existing SNN training methods when trained offline on real neural data while reducing the total memory usage by 92%. Additionally, it matches state-of-the-art accuracy in a closed loop environment, demonstrates adaptability when subjected to multiple types of neural input disruptions, and is capable of being trained online without any prior offline training.Significance.This work presents a neural decoding algorithm that can be trained rapidly in a closed loop setting. The algorithm increases the speed of acclimating a new user to the system and also can adapt to sudden changes in neural behavior with minimal disruption to the user.

Neuromorphic Photonics Circuits: Contemporary Review

Neuromorphic photonics is a cutting-edge fusion of neuroscience-inspired computing and photonics technology to overcome the constraints of conventional computing architectures. Its significance lies in the potential to transform information processing by mimicking the parallelism and efficiency of the human brain. Using optics and photonics principles, neuromorphic devices can execute intricate computations swiftly and with impressive energy efficiency. This innovation holds promise for advancing artificial intelligence and machine learning while addressing the limitations of traditional silicon-based computing. Neuromorphic photonics could herald a new era of computing that is more potent and draws inspiration from cognitive processes, leading to advancements in robotics, pattern recognition, and advanced data processing. This paper reviews the recent developments in neuromorphic photonic integrated circuits, applications, and current challenges.

Learnable Leakage and Onset-Spiking Self-Attention in SNNs with Local Error Signals

Spiking neural networks (SNNs) have garnered significant attention due to their computational patterns resembling biological neural networks. However, when it comes to deep SNNs, how to focus on critical information effectively and achieve a balanced feature transformation both temporally and spatially becomes a critical challenge. To address these challenges, our research is centered around two aspects: structure and strategy. Structurally, we optimize the leaky integrate-and-fire (LIF) neuron to enable the leakage coefficient to be learnable, thus making it better suited for contemporary applications. Furthermore, the self-attention mechanism is introduced at the initial time step to ensure improved focus and processing. Strategically, we propose a new normalization method anchored on the learnable leakage coefficient (LLC) and introduce a local loss signal strategy to enhance the SNN's training efficiency and adaptability. The effectiveness and performance of our proposed methods are validated on the MNIST, FashionMNIST, and CIFAR-10 datasets. Experimental results show that our model presents a superior, high-accuracy performance in just eight time steps. In summary, our research provides fresh insights into the structure and strategy of SNNs, paving the way for their efficient and robust application in practical scenarios.

Event Transformer +. A Multi-Purpose Solution for Efficient Event Data Processing

Event cameras record sparse illumination changes with high temporal resolution and high dynamic range. Thanks to their sparse recording and low consumption, they are increasingly used in applications such as AR/VR and autonomous driving. Current top-performing methods often ignore specific event-data properties, leading to the development of generic but computationally expensive algorithms, while event-aware methods do not perform as well. We propose Event Transformer +, that improves our seminal work EvT with a refined patch-based event representation and a more robust backbone to achieve more accurate results, while still benefiting from event-data sparsity to increase its efficiency. Additionally, we show how our system can work with different data modalities and propose specific output heads, for event-stream classification (i.e. action recognition) and per-pixel predictions (dense depth estimation). Evaluation results show better performance to the state-of-the-art while requiring minimal computation resources, both on GPU and CPU.

Online Spatio-Temporal Learning in Deep Neural Networks

Biological neural networks are equipped with an inherent capability to continuously adapt through online learning. This aspect remains in stark contrast to learning with error backpropagation through time (BPTT) that involves offline computation of the gradients due to the need to unroll the network through time. Here, we present an alternative online learning algorithm ic framework for deep recurrent neural networks (RNNs) and spiking neural networks (SNNs), called online spatio-temporal learning (OSTL). It is based on insights from biology and proposes the clear separation of spatial and temporal gradient components. For shallow SNNs, OSTL is gradient equivalent to BPTT enabling for the first time online training of SNNs with BPTT-equivalent gradients. In addition, the proposed formulation unveils a class of SNN architectures trainable online at low time complexity. Moreover, we extend OSTL to a generic form, applicable to a wide range of network architectures, including networks comprising long short-term memory (LSTM) and gated recurrent units (GRUs). We demonstrate the operation of our algorithm ic framework on various tasks from language modeling to speech recognition and obtain results on par with the BPTT baselines.

SpikingJelly: An open-source machine learning infrastructure platform for spike-based intelligence

Spiking neural networks (SNNs) aim to realize brain-inspired intelligence on neuromorphic chips with high energy efficiency by introducing neural dynamics and spike properties. As the emerging spiking deep learning paradigm attracts increasing interest, traditional programming frameworks cannot meet the demands of the automatic differentiation, parallel computation acceleration, and high integration of processing neuromorphic datasets and deployment. In this work, we present the SpikingJelly framework to address the aforementioned dilemma. We contribute a full-stack toolkit for preprocessing neuromorphic datasets, building deep SNNs, optimizing their parameters, and deploying SNNs on neuromorphic chips. Compared to existing methods, the training of deep SNNs can be accelerated 11×, and the superior extensibility and flexibility of SpikingJelly enable users to accelerate custom models at low costs through multilevel inheritance and semiautomatic code generation. SpikingJelly paves the way for synthesizing truly energy-efficient SNN-based machine intelligence systems, which will enrich the ecology of neuromorphic computing.

First-spike coding promotes accurate and efficient spiking neural networks for discrete events with rich temporal structures

Spiking neural networks (SNNs) are well-suited to process asynchronous event-based data. Most of the existing SNNs use rate-coding schemes that focus on firing rate (FR), and so they generally ignore the spike timing in events. On the contrary, methods based on temporal coding, particularly time-to-first-spike (TTFS) coding, can be accurate and efficient but they are difficult to train. Currently, there is limited research on applying TTFS coding to real events, since traditional TTFS-based methods impose one-spike constraint, which is not realistic for event-based data. In this study, we present a novel decision-making strategy based on first-spike (FS) coding that encodes FS timings of the output neurons to investigate the role of the first-spike timing in classifying real-world event sequences with complex temporal structures. To achieve FS coding, we propose a novel surrogate gradient learning method for discrete spike trains. In the forward pass, output spikes are encoded into discrete times to generate FS times. In the backpropagation, we develop an error assignment method that propagates error from FS times to spikes through a Gaussian window, and then supervised learning for spikes is implemented through a surrogate gradient approach. Additional strategies are introduced to facilitate the training of FS timings, such as adding empty sequences and employing different parameters for different layers. We make a comprehensive comparison between FS and FR coding in the experiments. Our results show that FS coding achieves comparable accuracy to FR coding while leading to superior energy efficiency and distinct neuronal dynamics on data sequences with very rich temporal structures. Additionally, a longer time delay in the first spike leads to higher accuracy, indicating important information is encoded in the timing of the first spike.

CQ + Training: Minimizing Accuracy Loss in Conversion From Convolutional Neural Networks to Spiking Neural Networks

Spiking neural networks (SNNs) are attractive for energy-constrained use-cases due to their binarized activation, eliminating the need for weight multiplication. However, its lag in accuracy compared to traditional convolutional network networks (CNNs) has limited its deployment. In this paper, we propose CQ+ training (extended "clamped" and "quantized" training), an SNN-compatible CNN training algorithm that achieves state-of-the-art accuracy for both CIFAR-10 and CIFAR-100 datasets. Using a 7-layer modified VGG model (VGG-*), we achieved 95.06% accuracy on the CIFAR-10 dataset for equivalent SNNs. The accuracy drop from converting the CNN solution to an SNN is only 0.09% when using a time step of 600. To reduce the latency, we propose a parameterized input encoding method and a threshold training method, which further reduces the time window size to 64 while still achieving an accuracy of 94.09%. For the CIFAR-100 dataset, we achieved an accuracy of 77.27% using the same VGG-* structure and a time window of 500. We also demonstrate the transformation of popular CNNs, including ResNet (basic, bottleneck, and shortcut block), MobileNet v1/2, and Densenet, to SNNs with near-zero conversion accuracy loss and a time window size smaller than 60. The framework was developed in PyTorch and is publicly available.

Complex Oxides for Brain-Inspired Computing: A Review

The fields of brain-inspired computing, robotics, and, more broadly, artificial intelligence (AI) seek to implement knowledge gleaned from the natural world into human-designed electronics and machines. In this review, the opportunities presented by complex oxides, a class of electronic ceramic materials whose properties can be elegantly tuned by doping, electron interactions, and a variety of external stimuli near room temperature, are discussed. The review begins with a discussion of natural intelligence at the elementary level in the nervous system, followed by collective intelligence and learning at the animal colony level mediated by social interactions. An important aspect highlighted is the vast spatial and temporal scales involved in learning and memory. The focus then turns to collective phenomena, such as metal-to-insulator transitions (MITs), ferroelectricity, and related examples, to highlight recent demonstrations of artificial neurons, synapses, and circuits and their learning. First-principles theoretical treatments of the electronic structure, and in situ synchrotron spectroscopy of operating devices are then discussed. The implementation of the experimental characteristics into neural networks and algorithm design is then revewed. Finally, outstanding materials challenges that require a microscopic understanding of the physical mechanisms, which will be essential for advancing the frontiers of neuromorphic computing, are highlighted.

Deep Spike Learning With Local Classifiers

Backpropagation has been successfully generalized to optimize deep spiking neural networks (SNNs), where, nevertheless, gradients need to be propagated back through all layers, resulting in a massive consumption of computing resources and an obstacle to the parallelization of training. A biologically motivated scheme of local learning provides an alternative to efficiently train deep networks but often suffers a low performance of accuracy on practical tasks. Thus, how to train deep SNNs with the local learning scheme to achieve both efficient and accurate performance still remains an important challenge. In this study, we focus on a supervised local learning scheme where each layer is independently optimized with an auxiliary classifier. Accordingly, we first propose a spike-based efficient local learning rule by only considering the direct dependencies in the current time. We then propose two variants that additionally incorporate temporal dependencies through a backward and forward process, respectively. The effectiveness and performance of our proposed methods are extensively evaluated with six mainstream datasets. Experimental results show that our methods can successfully scale up to large networks and substantially outperform the spike-based local learning baselines on all studied benchmarks. Our results also reveal that gradients with temporal dependencies are essential for high performance on temporal tasks, while they have negligible effects on rate-based tasks. Our work is significant as it brings the performance of spike-based local learning to a new level with the computational benefits being retained.

Critically synchronized brain waves form an effective, robust and flexible basis for human memory and learning

The effectiveness, robustness, and flexibility of memory and learning constitute the very essence of human natural intelligence, cognition, and consciousness. However, currently accepted views on these subjects have, to date, been put forth without any basis on a true physical theory of how the brain communicates internally via its electrical signals. This lack of a solid theoretical framework has implications not only for our understanding of how the brain works, but also for wide range of computational models developed from the standard orthodox view of brain neuronal organization and brain network derived functioning based on the Hodgkin-Huxley ad-hoc circuit analogies that have produced a multitude of Artificial, Recurrent, Convolution, Spiking, etc., Neural Networks (ARCSe NNs) that have in turn led to the standard algorithms that form the basis of artificial intelligence (AI) and machine learning (ML) methods. Our hypothesis, based upon our recently developed physical model of weakly evanescent brain wave propagation (WETCOW) is that, contrary to the current orthodox model that brain neurons just integrate and fire under accompaniment of slow leaking, they can instead perform much more sophisticated tasks of efficient coherent synchronization/desynchronization guided by the collective influence of propagating nonlinear near critical brain waves, the waves that currently assumed to be nothing but inconsequential subthreshold noise. In this paper we highlight the learning and memory capabilities of our WETCOW framework and then apply it to the specific application of AI/ML and Neural Networks. We demonstrate that the learning inspired by these critically synchronized brain waves is shallow, yet its timing and accuracy outperforms deep ARCSe counterparts on standard test datasets. These results have implications for both our understanding of brain function and for the wide range of AI/ML applications.

Presynaptic spike-driven plasticity based on eligibility trace for on-chip learning system

Introduction

Recurrent spiking neural network (RSNN) performs excellently in spatio-temporal learning with backpropagation through time (BPTT) algorithm. But the requirement of computation and memory in BPTT makes it hard to realize an on-chip learning system based on RSNN. In this paper, we aim to realize a high-efficient RSNN learning system on field programmable gate array (FPGA).

Methods

A presynaptic spike-driven plasticity architecture based on eligibility trace is implemented to reduce the resource consumption. The RSNN with leaky integrate-and-fire (LIF) and adaptive LIF (ALIF) models is implemented on FPGA based on presynaptic spike-driven architecture. In this architecture, the eligibility trace gated by a learning signal is used to optimize synaptic weights without unfolding the network through time. When a presynaptic spike occurs, the eligibility trace is calculated based on its latest timestamp and drives synapses to update their weights. Only the latest timestamps of presynaptic spikes are required to be stored in buffers to calculate eligibility traces.

Results

We show the implementation of this architecture on FPGA and test it with two experiments. With the presynaptic spike-driven architecture, the resource consumptions, including look-up tables (LUTs) and registers, and dynamic power consumption of synaptic modules in the on-chip learning system are greatly reduced. The experiment results and compilation results show that the buffer size of the on-chip learning system is reduced and the RSNNs implemented on FPGA exhibit high efficiency in resources and energy while accurately solving tasks.

Discussion

This study provides a solution to the problem of data congestion in the buffer of large-scale learning systems.

Efficient training of spiking neural networks with temporally-truncated local backpropagation through time

Directly training spiking neural networks (SNNs) has remained challenging due to complex neural dynamics and intrinsic non-differentiability in firing functions. The well-known backpropagation through time (BPTT) algorithm proposed to train SNNs suffers from large memory footprint and prohibits backward and update unlocking, making it impossible to exploit the potential of locally-supervised training methods. This work proposes an efficient and direct training algorithm for SNNs that integrates a locally-supervised training method with a temporally-truncated BPTT algorithm. The proposed algorithm explores both temporal and spatial locality in BPTT and contributes to significant reduction in computational cost including GPU memory utilization, main memory access and arithmetic operations. We thoroughly explore the design space concerning temporal truncation length and local training block size and benchmark their impact on classification accuracy of different networks running different types of tasks. The results reveal that temporal truncation has a negative effect on the accuracy of classifying frame-based datasets, but leads to improvement in accuracy on event-based datasets. In spite of resulting information loss, local training is capable of alleviating overfitting. The combined effect of temporal truncation and local training can lead to the slowdown of accuracy drop and even improvement in accuracy. In addition, training deep SNNs' models such as AlexNet classifying CIFAR10-DVS dataset leads to 7.26% increase in accuracy, 89.94% reduction in GPU memory, 10.79% reduction in memory access, and 99.64% reduction in MAC operations compared to the standard end-to-end BPTT. Thus, the proposed method has shown high potential to enable fast and energy-efficient on-chip training for real-time learning at the edge.

STSC-SNN: Spatio-Temporal Synaptic Connection with temporal convolution and attention for spiking neural networks

Spiking neural networks (SNNs), as one of the algorithmic models in neuromorphic computing, have gained a great deal of research attention owing to temporal information processing capability, low power consumption, and high biological plausibility. The potential to efficiently extract spatio-temporal features makes it suitable for processing event streams. However, existing synaptic structures in SNNs are almost full-connections or spatial 2D convolution, neither of which can extract temporal dependencies adequately. In this work, we take inspiration from biological synapses and propose a Spatio-Temporal Synaptic Connection SNN (STSC-SNN) model to enhance the spatio-temporal receptive fields of synaptic connections, thereby establishing temporal dependencies across layers. Specifically, we incorporate temporal convolution and attention mechanisms to implement synaptic filtering and gating functions. We show that endowing synaptic models with temporal dependencies can improve the performance of SNNs on classification tasks. In addition, we investigate the impact of performance via varied spatial-temporal receptive fields and reevaluate the temporal modules in SNNs. Our approach is tested on neuromorphic datasets, including DVS128 Gesture (gesture recognition), N-MNIST, CIFAR10-DVS (image classification), and SHD (speech digit recognition). The results show that the proposed model outperforms the state-of-the-art accuracy on nearly all datasets.

A Co-Designed Neuromorphic Chip With Compact (17.9K F2) and Weak Neuron Number-Dependent Neuron/Synapse Modules

Many efforts have been made to improve the neuron integration efficiency on neuromorphic chips, such as using emerging memory devices and shrinking CMOS technology nodes. However, in the fully connected (FC) neuromorphic core, increasing the number of neurons will lead to a square increase in synapse & dendrite costs and a high-slope linear increase in soma costs, resulting in an explosive growth of core hardware costs. We propose a co-designed neuromorphic core (SRCcore) based on the quantized spiking neural network (SNN) technology and compact chip design methodology. The cost of the neuron/synapse module in SRCcore weakly depends on the neuron number, which effectively relieves the growth pressure of the core area caused by increasing the neuron number. In the proposed BICS chip based on SRCcore, although the neuron/synapse module implements 1∼16 times of neurons and 1∼66 times of synapses, it only costs an area of 1.79 × 10⁷ F², which is 7.9%∼38.6% of that in previous works. Based on the weight quantization strategy matched with SRCcore, quantized SNNs achieve 0.05%∼2.19% higher accuracy than previous works, thus supporting the design and application of SRCcore. Finally, a cross-modeling application is demonstrated based on the chip. We hope this work will accelerate the development of cortical-scale neuromorphic systems.

E-prop on SpiNNaker 2: Exploring online learning in spiking RNNs on neuromorphic hardware

Introduction

In recent years, the application of deep learning models at the edge has gained attention. Typically, artificial neural networks (ANNs) are trained on graphics processing units (GPUs) and optimized for efficient execution on edge devices. Training ANNs directly at the edge is the next step with many applications such as the adaptation of models to specific situations like changes in environmental settings or optimization for individuals, e.g., optimization for speakers for speech processing. Also, local training can preserve privacy. Over the last few years, many algorithms have been developed to reduce memory footprint and computation.

Methods

A specific challenge to train recurrent neural networks (RNNs) for processing sequential data is the need for the Back Propagation Through Time (BPTT) algorithm to store the network state of all time steps. This limitation is resolved by the biologically-inspired E-prop approach for training Spiking Recurrent Neural Networks (SRNNs). We implement the E-prop algorithm on a prototype of the SpiNNaker 2 neuromorphic system. A parallelization strategy is developed to split and train networks on the ARM cores of SpiNNaker 2 to make efficient use of both memory and compute resources. We trained an SRNN from scratch on SpiNNaker 2 in real-time on the Google Speech Command dataset for keyword spotting.

Result

We achieved an accuracy of 91.12% while requiring only 680 KB of memory for training the network with 25 K weights. Compared to other spiking neural networks with equal or better accuracy, our work is significantly more memory-efficient.

Discussion

In addition, we performed a memory and time profiling of the E-prop algorithm. This is used on the one hand to discuss whether E-prop or BPTT is better suited for training a model at the edge and on the other hand to explore architecture modifications to SpiNNaker 2 to speed up online learning. Finally, energy estimations predict that the SRNN can be trained on SpiNNaker2 with 12 times less energy than using a NVIDIA V100 GPU.

Bayesian continual learning via spiking neural networks

Among the main features of biological intelligence are energy efficiency, capacity for continual adaptation, and risk management via uncertainty quantification. Neuromorphic engineering has been thus far mostly driven by the goal of implementing energy-efficient machines that take inspiration from the time-based computing paradigm of biological brains. In this paper, we take steps toward the design of neuromorphic systems that are capable of adaptation to changing learning tasks, while producing well-calibrated uncertainty quantification estimates. To this end, we derive online learning rules for spiking neural networks (SNNs) within a Bayesian continual learning framework. In it, each synaptic weight is represented by parameters that quantify the current epistemic uncertainty resulting from prior knowledge and observed data. The proposed online rules update the distribution parameters in a streaming fashion as data are observed. We instantiate the proposed approach for both real-valued and binary synaptic weights. Experimental results using Intel's Lava platform show the merits of Bayesian over frequentist learning in terms of capacity for adaptation and uncertainty quantification.

A Parallel Spiking Neural Network Based on Adaptive Lateral Inhibition Mechanism for Objective Recognition

Spiking neural network (SNN) has attracted extensive attention in the field of machine learning because of its biological interpretability and low power consumption. However, the accuracy of pattern recognition cannot completely surpass deep neural networks (DNNs). The main reason is that the inherent nondifferentiability of spiking neurons makes SNN unable to be trained directly by the gradient descent algorithm, and there is also no unified training algorithm for SNN. Inspired by the biological vision system, this paper proposes a parallel convolution SNN structure combined with an adaptive lateral inhibition mechanism. And, a way of dynamically evolving the time constant with the training of SNN is proposed to ensure the diversity of neurons. This paper verifies the effectiveness of the proposed methods on static datasets and neuromorphic datasets and extends it to the recognition of breast tumors. Experimental results show that the SNN has obvious advantages in dynamical datasets. For breast tumors, it is also an edge-based task, because the edge of a medical image contains the most important information in the image. This kind of information can provide great help for the noninvasive and accurate diagnosis of diseases. The Experimental results show that the proposed method is very close to the recognition results of DNNs on static datasets, and its performance on neuromorphic datasets exceeds that of DNNs.

Highly efficient neuromorphic learning system of spiking neural network with multi-compartment leaky integrate-and-fire neurons

A spiking neural network (SNN) is considered a high-performance learning system that matches the digital circuits and presents higher efficiency due to the architecture and computation of spiking neurons. While implementing a SNN on a field-programmable gate array (FPGA), the gradient back-propagation through layers consumes a surprising number of resources. In this paper, we aim to realize an efficient architecture of SNN on the FPGA to reduce resource and power consumption. The multi-compartment leaky integrate-and-fire (MLIF) model is used to convert spike trains to the plateau potential in dendrites. We accumulate the potential in the apical dendrite during the training period. The average of this accumulative result is the dendritic plateau potential and is used to guide the updates of synaptic weights. Based on this architecture, the SNN is implemented on FPGA efficiently. In the implementation of a neuromorphic learning system, the shift multiplier (shift MUL) module and piecewise linear (PWL) algorithm are used to replace multipliers and complex nonlinear functions to match the digital circuits. The neuromorphic learning system is constructed with resources on FPGA without dataflow between on-chip and off-chip memories. Our neuromorphic learning system performs with higher resource utilization and power efficiency than previous on-chip learning systems.

Toward the Optimal Design and FPGA Implementation of Spiking Neural Networks

The performance of a biologically plausible spiking neural network (SNN) largely depends on the model parameters and neural dynamics. This article proposes a parameter optimization scheme for improving the performance of a biologically plausible SNN and a parallel on-field-programmable gate array (FPGA) online learning neuromorphic platform for the digital implementation based on two numerical methods, namely, the Euler and third-order Runge-Kutta (RK3) methods. The optimization scheme explores the impact of biological time constants on information transmission in the SNN and improves the convergence rate of the SNN on digit recognition with a suitable choice of the time constants. The parallel digital implementation leads to a significant speedup over software simulation on a general-purpose CPU. The parallel implementation with the Euler method enables around 180× ( 20× ) training (inference) speedup over a Pytorch-based SNN simulation on CPU. Moreover, compared with previous work, our parallel implementation shows more than 300× ( 240× ) improvement on speed and 180× ( 250× ) reduction in energy consumption for training (inference). In addition, due to the high-order accuracy, the RK3 method is demonstrated to gain 2× training speedup over the Euler method, which makes it suitable for online training in real-time applications.

Memory-inspired spiking hyperdimensional network for robust online learning

Recently, brain-inspired computing models have shown great potential to outperform today's deep learning solutions in terms of robustness and energy efficiency. Particularly, Spiking Neural Networks (SNNs) and HyperDimensional Computing (HDC) have shown promising results in enabling efficient and robust cognitive learning. Despite the success, these two brain-inspired models have different strengths. While SNN mimics the physical properties of the human brain, HDC models the brain on a more abstract and functional level. Their design philosophies demonstrate complementary patterns that motivate their combination. With the help of the classical psychological model on memory, we propose SpikeHD, the first framework that fundamentally combines Spiking neural network and hyperdimensional computing. SpikeHD generates a scalable and strong cognitive learning system that better mimics brain functionality. SpikeHD exploits spiking neural networks to extract low-level features by preserving the spatial and temporal correlation of raw event-based spike data. Then, it utilizes HDC to operate over SNN output by mapping the signal into high-dimensional space, learning the abstract information, and classifying the data. Our extensive evaluation on a set of benchmark classification problems shows that SpikeHD provides the following benefit compared to SNN architecture: (1) significantly enhance learning capability by exploiting two-stage information processing, (2) enables substantial robustness to noise and failure, and (3) reduces the network size and required parameters to learn complex information.

Multisample Online Learning for Probabilistic Spiking Neural Networks

Spiking neural networks (SNNs) capture some of the efficiency of biological brains for inference and learning via the dynamic, online, and event-driven processing of binary time series. Most existing learning algorithms for SNNs are based on deterministic neuronal models, such as leaky integrate-and-fire, and rely on heuristic approximations of backpropagation through time that enforces constraints such as locality. In contrast, probabilistic SNN models can be trained directly via principled online, local, and update rules that have proven to be particularly effective for resource-constrained systems. This article investigates another advantage of probabilistic SNNs, namely, their capacity to generate independent outputs when queried over the same input. It is shown that the multiple generated output samples can be used during inference to robustify decisions and to quantify uncertainty-a feature that deterministic SNN models cannot provide. Furthermore, they can be leveraged for training in order to obtain more accurate statistical estimates of the log-loss training criterion and its gradient. Specifically, this article introduces an online learning rule based on generalized expectation-maximization (GEM) that follows a three-factor form with global learning signals and is referred to as GEM-SNN. Experimental results on structured output memorization and classification on a standard neuromorphic dataset demonstrate significant improvements in terms of log-likelihood, accuracy, and calibration when increasing the number of samples used for inference and training.

Backpropagation With Sparsity Regularization for Spiking Neural Network Learning

The spiking neural network (SNN) is a possible pathway for low-power and energy-efficient processing and computing exploiting spiking-driven and sparsity features of biological systems. This article proposes a sparsity-driven SNN learning algorithm, namely backpropagation with sparsity regularization (BPSR), aiming to achieve improved spiking and synaptic sparsity. Backpropagation incorporating spiking regularization is utilized to minimize the spiking firing rate with guaranteed accuracy. Backpropagation realizes the temporal information capture and extends to the spiking recurrent layer to support brain-like structure learning. The rewiring mechanism with synaptic regularization is suggested to further mitigate the redundancy of the network structure. Rewiring based on weight and gradient regulates the pruning and growth of synapses. Experimental results demonstrate that the network learned by BPSR has synaptic sparsity and is highly similar to the biological system. It not only balances the accuracy and firing rate, but also facilitates SNN learning by suppressing the information redundancy. We evaluate the proposed BPSR on the visual dataset MNIST, N-MNIST, and CIFAR10, and further test it on the sensor dataset MIT-BIH and gas sensor. Results bespeak that our algorithm achieves comparable or superior accuracy compared to related works, with sparse spikes and synapses.

Multilayer Photonic Spiking Neural Networks: Generalized Supervised Learning Algorithm and Network Optimization

We propose a generalized supervised learning algorithm for multilayer photonic spiking neural networks (SNNs) by combining the spike-timing dependent plasticity (STDP) rule and the gradient descent mechanism. A vertical-cavity surface-emitting laser with an embedded saturable absorber (VCSEL-SA) is employed as a photonic leaky-integrate-and-fire (LIF) neuron. The temporal coding strategy is employed to transform information into the precise firing time. With the modified supervised learning algorithm, the trained multilayer photonic SNN successfully solves the XOR problem and performs well on the Iris and Wisconsin breast cancer datasets. This indicates that a generalized supervised learning algorithm is realized for multilayer photonic SNN. In addition, network optimization is performed by considering different network sizes.

Training Deep Convolutional Spiking Neural Networks With Spike Probabilistic Global Pooling

Recent work on spiking neural networks (SNNs) has focused on achieving deep architectures. They commonly use backpropagation (BP) to train SNNs directly, which allows SNNs to go deeper and achieve higher performance. However, the BP training procedure is computing intensive and complicated by many trainable parameters. Inspired by global pooling in convolutional neural networks (CNNs), we present the spike probabilistic global pooling (SPGP) method based on a probability function for training deep convolutional SNNs. It aims to remove the difficult of too many trainable parameters brought by multiple layers in the training process, which can reduce the risk of overfitting and get better performance for deep SNNs (DSNNs). We use the discrete leaky-integrate-fire model and the spatiotemporal BP algorithm for training DSNNs directly. As a result, our model trained with the SPGP method achieves competitive performance compared to the existing DSNNs on image and neuromorphic data sets while minimizing the number of trainable parameters. In addition, the proposed SPGP method shows its effectiveness in performance improvement, convergence, and generalization ability.

Brain-inspired global-local learning incorporated with neuromorphic computing

There are two principle approaches for learning in artificial intelligence: error-driven global learning and neuroscience-oriented local learning. Integrating them into one network may provide complementary learning capabilities for versatile learning scenarios. At the same time, neuromorphic computing holds great promise, but still needs plenty of useful algorithms and algorithm-hardware co-designs to fully exploit its advantages. Here, we present a neuromorphic global-local synergic learning model by introducing a brain-inspired meta-learning paradigm and a differentiable spiking model incorporating neuronal dynamics and synaptic plasticity. It can meta-learn local plasticity and receive top-down supervision information for multiscale learning. We demonstrate the advantages of this model in multiple different tasks, including few-shot learning, continual learning, and fault-tolerance learning in neuromorphic vision sensors. It achieves significantly higher performance than single-learning methods. We further implement the model in the Tianjic neuromorphic platform by exploiting algorithm-hardware co-designs and prove that the model can fully utilize neuromorphic many-core architecture to develop hybrid computation paradigm.

Opportunities for neuromorphic computing algorithms and applications

Neuromorphic computing technologies will be important for the future of computing, but much of the work in neuromorphic computing has focused on hardware development. Here, we review recent results in neuromorphic computing algorithms and applications. We highlight characteristics of neuromorphic computing technologies that make them attractive for the future of computing and we discuss opportunities for future development of algorithms and applications on these systems.

A flexible capacitive photoreceptor for the biomimetic retina

Neuromorphic vision sensors have been extremely beneficial in developing energy-efficient intelligent systems for robotics and privacy-preserving security applications. There is a dire need for devices to mimic the retina's photoreceptors that encode the light illumination into a sequence of spikes to develop such sensors. Herein, we develop a hybrid perovskite-based flexible photoreceptor whose capacitance changes proportionally to the light intensity mimicking the retina's rod cells, paving the way for developing an efficient artificial retina network. The proposed device constitutes a hybrid nanocomposite of perovskites (methyl-ammonium lead bromide) and the ferroelectric terpolymer (polyvinylidene fluoride trifluoroethylene-chlorofluoroethylene). A metal-insulator-metal type capacitor with the prepared composite exhibits the unique and photosensitive capacitive behavior at various light intensities in the visible light spectrum. The proposed photoreceptor mimics the spectral sensitivity curve of human photopic vision. The hybrid nanocomposite is stable in ambient air for 129 weeks, with no observable degradation of the composite due to the encapsulation of hybrid perovskites in the hydrophobic polymer. The functionality of the proposed photoreceptor to recognize handwritten digits (MNIST) dataset using an unsupervised trained spiking neural network with 72.05% recognition accuracy is demonstrated. This demonstration proves the potential of the proposed sensor for neuromorphic vision applications.

Supervised training of spiking neural networks for robust deployment on mixed-signal neuromorphic processors

Mixed-signal analog/digital circuits emulate spiking neurons and synapses with extremely high energy efficiency, an approach known as "neuromorphic engineering". However, analog circuits are sensitive to process-induced variation among transistors in a chip ("device mismatch"). For neuromorphic implementation of Spiking Neural Networks (SNNs), mismatch causes parameter variation between identically-configured neurons and synapses. Each chip exhibits a different distribution of neural parameters, causing deployed networks to respond differently between chips. Current solutions to mitigate mismatch based on per-chip calibration or on-chip learning entail increased design complexity, area and cost, making deployment of neuromorphic devices expensive and difficult. Here we present a supervised learning approach that produces SNNs with high robustness to mismatch and other common sources of noise. Our method trains SNNs to perform temporal classification tasks by mimicking a pre-trained dynamical system, using a local learning rule from non-linear control theory. We demonstrate our method on two tasks requiring temporal memory, and measure the robustness of our approach to several forms of noise and mismatch. We show that our approach is more robust than common alternatives for training SNNs. Our method provides robust deployment of pre-trained networks on mixed-signal neuromorphic hardware, without requiring per-device training or calibration.

BlocTrain: Block-Wise Conditional Training and Inference for Efficient Spike-Based Deep Learning

Spiking neural networks (SNNs), with their inherent capability to learn sparse spike-based input representations over time, offer a promising solution for enabling the next generation of intelligent autonomous systems. Nevertheless, end-to-end training of deep SNNs is both compute- and memory-intensive because of the need to backpropagate error gradients through time. We propose BlocTrain, which is a scalable and complexity-aware incremental algorithm for memory-efficient training of deep SNNs. We divide a deep SNN into blocks, where each block consists of few convolutional layers followed by a classifier. We train the blocks sequentially using local errors from the classifier. Once a given block is trained, our algorithm dynamically figures out easy vs. hard classes using the class-wise accuracy, and trains the deeper block only on the hard class inputs. In addition, we also incorporate a hard class detector (HCD) per block that is used during inference to exit early for the easy class inputs and activate the deeper blocks only for the hard class inputs. We trained ResNet-9 SNN divided into three blocks, using BlocTrain, on CIFAR-10 and obtained 86.4% accuracy, which is achieved with up to 2.95× lower memory requirement during the course of training, and 1.89× compute efficiency per inference (due to early exit strategy) with 1.45× memory overhead (primarily due to classifier weights) compared to end-to-end network. We also trained ResNet-11, divided into four blocks, on CIFAR-100 and obtained 58.21% accuracy, which is one of the first reported accuracy for SNN trained entirely with spike-based backpropagation on CIFAR-100.

μBrain: An Event-Driven and Fully Synthesizable Architecture for Spiking Neural Networks

The development of brain-inspired neuromorphic computing architectures as a paradigm for Artificial Intelligence (AI) at the edge is a candidate solution that can meet strict energy and cost reduction constraints in the Internet of Things (IoT) application areas. Toward this goal, we present μBrain: the first digital yet fully event-driven without clock architecture, with co-located memory and processing capability that exploits event-based processing to reduce an always-on system's overall energy consumption (μW dynamic operation). The chip area in a 40 nm Complementary Metal Oxide Semiconductor (CMOS) digital technology is 2.82 mm² including pads (without pads 1.42 mm²). This small area footprint enables μBrain integration in re-trainable sensor ICs to perform various signal processing tasks, such as data preprocessing, dimensionality reduction, feature selection, and application-specific inference. We present an instantiation of the μBrain architecture in a 40 nm CMOS digital chip and demonstrate its efficiency in a radar-based gesture classification with a power consumption of 70 μW and energy consumption of 340 nJ per classification. As a digital architecture, μBrain is fully synthesizable and lends to a fast development-to-deployment cycle in Application-Specific Integrated Circuits (ASIC). To the best of our knowledge, μBrain is the first tiny-scale digital, spike-based, fully parallel, non-Von-Neumann architecture (without schedules, clocks, nor state machines). For these reasons, μBrain is ultra-low-power and offers software-to-hardware fidelity. μBrain enables always-on neuromorphic computing in IoT sensor nodes that require running on battery power for years.

Is Neuromorphic MNIST Neuromorphic? Analyzing the Discriminative Power of Neuromorphic Datasets in the Time Domain

A major characteristic of spiking neural networks (SNNs) over conventional artificial neural networks (ANNs) is their ability to spike, enabling them to use spike timing for coding and efficient computing. In this paper, we assess if neuromorphic datasets recorded from static images are able to evaluate the ability of SNNs to use spike timings in their calculations. We have analyzed N-MNIST, N-Caltech101 and DvsGesture along these lines, but focus our study on N-MNIST. First we evaluate if additional information is encoded in the time domain in a neuromorphic dataset. We show that an ANN trained with backpropagation on frame-based versions of N-MNIST and N-Caltech101 images achieve 99.23 and 78.01% accuracy. These are comparable to the state of the art-showing that an algorithm that purely works on spatial data can classify these datasets. Second we compare N-MNIST and DvsGesture on two STDP algorithms, RD-STDP, that can classify only spatial data, and STDP-tempotron that classifies spatiotemporal data. We demonstrate that RD-STDP performs very well on N-MNIST, while STDP-tempotron performs better on DvsGesture. Since DvsGesture has a temporal dimension, it requires STDP-tempotron, while N-MNIST can be adequately classified by an algorithm that works on spatial data alone. This shows that precise spike timings are not important in N-MNIST. N-MNIST does not, therefore, highlight the ability of SNNs to classify temporal data. The conclusions of this paper open the question-what dataset can evaluate SNN ability to classify temporal data?

EqSpike: spike-driven equilibrium propagation for neuromorphic implementations

Finding spike-based learning algorithms that can be implemented within the local constraints of neuromorphic systems, while achieving high accuracy, remains a formidable challenge. Equilibrium propagation is a promising alternative to backpropagation as it only involves local computations, but hardware-oriented studies have so far focused on rate-based networks. In this work, we develop a spiking neural network algorithm called EqSpike, compatible with neuromorphic systems, which learns by equilibrium propagation. Through simulations, we obtain a test recognition accuracy of 97.6% on the MNIST handwritten digits dataset (Mixed National Institute of Standards and Technology), similar to rate-based equilibrium propagation, and comparing favorably to alternative learning techniques for spiking neural networks. We show that EqSpike implemented in silicon neuromorphic technology could reduce the energy consumption of inference and training, respectively, by three orders and two orders of magnitude compared to graphics processing units. Finally, we also show that during learning, EqSpike weight updates exhibit a form of spike-timing-dependent plasticity, highlighting a possible connection with biology.

Learning Without Feedback: Fixed Random Learning Signals Allow for Feedforward Training of Deep Neural Networks

While the backpropagation of error algorithm enables deep neural network training, it implies (i) bidirectional synaptic weight transport and (ii) update locking until the forward and backward passes are completed. Not only do these constraints preclude biological plausibility, but they also hinder the development of low-cost adaptive smart sensors at the edge, as they severely constrain memory accesses and entail buffering overhead. In this work, we show that the one-hot-encoded labels provided in supervised classification problems, denoted as targets, can be viewed as a proxy for the error sign. Therefore, their fixed random projections enable a layerwise feedforward training of the hidden layers, thus solving the weight transport and update locking problems while relaxing the computational and memory requirements. Based on these observations, we propose the direct random target projection (DRTP) algorithm and demonstrate that it provides a tradeoff between accuracy and computational cost that is suitable for adaptive edge computing devices.

Hardware Implementation of Deep Network Accelerators Towards Healthcare and Biomedical Applications

The advent of dedicated Deep Learning (DL) accelerators and neuromorphic processors has brought on new opportunities for applying both Deep and Spiking Neural Network (SNN) algorithms to healthcare and biomedical applications at the edge. This can facilitate the advancement of medical Internet of Things (IoT) systems and Point of Care (PoC) devices. In this paper, we provide a tutorial describing how various technologies including emerging memristive devices, Field Programmable Gate Arrays (FPGAs), and Complementary Metal Oxide Semiconductor (CMOS) can be used to develop efficient DL accelerators to solve a wide variety of diagnostic, pattern recognition, and signal processing problems in healthcare. Furthermore, we explore how spiking neuromorphic processors can complement their DL counterparts for processing biomedical signals. The tutorial is augmented with case studies of the vast literature on neural network and neuromorphic hardware as applied to the healthcare domain. We benchmark various hardware platforms by performing a sensor fusion signal processing task combining electromyography (EMG) signals with computer vision. Comparisons are made between dedicated neuromorphic processors and embedded AI accelerators in terms of inference latency and energy. Finally, we provide our analysis of the field and share a perspective on the advantages, disadvantages, challenges, and opportunities that various accelerators and neuromorphic processors introduce to healthcare and biomedical domains.

Integration and Co-design of Memristive Devices and Algorithms for Artificial Intelligence

Memristive devices share remarkable similarities to biological synapses, dendrites, and neurons at both the physical mechanism level and unit functionality level, making the memristive approach to neuromorphic computing a promising technology for future artificial intelligence. However, these similarities do not directly transfer to the success of efficient computation without device and algorithm co-designs and optimizations. Contemporary deep learning algorithms demand the memristive artificial synapses to ideally possess analog weighting and linear weight-update behavior, requiring substantial device-level and circuit-level optimization. Such co-design and optimization have been the main focus of memristive neuromorphic engineering, which often abandons the “non-ideal” behaviors of memristive devices, although many of them resemble what have been observed in biological components. Novel brain-inspired algorithms are being proposed to utilize such behaviors as unique features to further enhance the efficiency and intelligence of neuromorphic computing, which calls for collaborations among electrical engineers, computing scientists, and neuroscientists.

A solution to the learning dilemma for recurrent networks of spiking neurons

Recurrently connected networks of spiking neurons underlie the astounding information processing capabilities of the brain. Yet in spite of extensive research, how they can learn through synaptic plasticity to carry out complex network computations remains unclear. We argue that two pieces of this puzzle were provided by experimental data from neuroscience. A mathematical result tells us how these pieces need to be combined to enable biologically plausible online network learning through gradient descent, in particular deep reinforcement learning. This learning method-called e-prop-approaches the performance of backpropagation through time (BPTT), the best-known method for training recurrent neural networks in machine learning. In addition, it suggests a method for powerful on-chip learning in energy-efficient spike-based hardware for artificial intelligence.