Auto-Spikformer: Spikformer architecture search.

Introduction: The integration of self-attention mechanisms into Spiking Neural Networks (SNNs) has garnered considerable interest in the realm of advanced deep learning, primarily due to their biological properties. Recent advancements in SNN architecture, such as Spikformer, have demonstrated promising outcomes. However, we observe that Spikformer may exhibit excessive energy consumption, potentially attributable to redundant channels and blocks. Methods: To mitigate this issue, we propose a one-shot Spiking Transformer Architecture Search method, namely Auto-Spikformer. Auto-Spikformer extends the search space to include both transformer architecture and SNN inner parameters. We train and search the supernet based on weight entanglement, evolutionary search, and the proposed Discrete Spiking Parameters Search (DSPS) methods. Benefiting from these methods, the performance of subnets with weights inherited from the supernet without even retraining is comparable to the original Spikformer. Moreover, we propose a new fitness function aiming to find a Pareto optimal combination balancing energy consumption and accuracy. Results and discussion: Our experimental results demonstrate the effectiveness of Auto-Spikformer, which outperforms the original Spikformer and most CNN or ViT models with even fewer parameters and lower energy consumption.

A spatiotemporal energy model based on spiking neurons for human motion perception.

Inspired by the motion processing pathway, this paper proposes a bio-inspired feedforward spiking network model based on Hodgkin-Huxley neurons for human motion perception. The proposed network mimics the mechanisms of direction selectivity found in simple and complex cells of the primary visual cortex. Simple cells' receptive fields are modeled using Gabor energy filters, while complex cells' receptive fields are constructed by integrating the responses of simple cells in an energy model. To generate the motion map, the spiking output of the network integrates motion information encoded by the responses of complex cells with various preferred directions. Simulation results demonstrate that the spiking neuron-based network effectively replicates the directional selectivity operation of the visual cortex when presented with a sequence of time-varying images. We evaluate the proposed model against state-of-the-art spiking neuron-based motion detection models using publicly available datasets. The results highlight the model's capability to extract motion energy from diverse video sequences, akin to human visual motion perception models. Additionally, we showcase the application of the proposed model in motion segmentation tasks and compare its performance with state-of-the-art motion-based segmentation models using challenging video segmentation benchmarks. The results indicate competitive performance. The motion maps generated by the proposed model can be utilized for action recognition in input videos.

Predicting the temporal-dynamic trajectories of cortical neuronal responses in non-human primates based on deep spiking neural network.

Deep convolutional neural networks (CNNs) are commonly used as computational models for the primate ventral stream, while deep spiking neural networks (SNNs) incorporated with both the temporal and spatial spiking information still lack investigation. We compared performances of SNN and CNN in prediction of visual responses to the naturalistic stimuli in area V4, inferior temporal (IT), and orbitofrontal cortex (OFC). The accuracies based on SNN were significantly higher than that of CNN in prediction of temporal-dynamic trajectory and averaged firing rate of visual response in V4 and IT. The temporal dynamics were captured by SNN for neurons with diverse temporal profiles and category selectivities, and most sensitively captured around the time of peak responses for each brain region. Consistently, SNN activities showed significantly stronger correlations with IT, V4 and OFC responses. In SNN, correlations with neural activities were stronger for later time-step features than early time-step features. The temporal-dynamic prediction was also significantly improved by considering preceding neural activities during the prediction. Thus, our study demonstrated SNN as a powerful temporal-dynamic model for cortical responses to complex naturalistic stimuli.

SNN4Agents: a framework for developing energy-efficient embodied spiking neural networks for autonomous agents.

Recent trends have shown that autonomous agents, such as Autonomous Ground Vehicles (AGVs), Unmanned Aerial Vehicles (UAVs), and mobile robots, effectively improve human productivity in solving diverse tasks. However, since these agents are typically powered by portable batteries, they require extremely low power/energy consumption to operate in a long lifespan. To solve this challenge, neuromorphic computing has emerged as a promising solution, where bio-inspired Spiking Neural Networks (SNNs) use spikes from event-based cameras or data conversion pre-processing to perform sparse computations efficiently. However, the studies of SNN deployments for autonomous agents are still at an early stage. Hence, the optimization stages for enabling efficient embodied SNN deployments for autonomous agents have not been defined systematically. Toward this, we propose a novel framework called SNN4Agents that consists of a set of optimization techniques for designing energy-efficient embodied SNNs targeting autonomous agent applications. Our SNN4Agents employs weight quantization, timestep reduction, and attention window reduction to jointly improve the energy efficiency, reduce the memory footprint, optimize the processing latency, while maintaining high accuracy. In the evaluation, we investigate use cases of event-based car recognition, and explore the trade-offs among accuracy, latency, memory, and energy consumption. The experimental results show that our proposed framework can maintain high accuracy (i.e., 84.12% accuracy) with 68.75% memory saving, 3.58x speed-up, and 4.03x energy efficiency improvement as compared to the state-of-the-art work for the NCARS dataset. In this manner, our SNN4Agents framework paves the way toward enabling energy-efficient embodied SNN deployments for autonomous agents.

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.

Composing recurrent spiking neural networks using locally-recurrent motifs and risk-mitigating architectural optimization.

In neural circuits, recurrent connectivity plays a crucial role in network function and stability. However, existing recurrent spiking neural networks (RSNNs) are often constructed by random connections without optimization. While RSNNs can produce rich dynamics that are critical for memory formation and learning, systemic architectural optimization of RSNNs is still an open challenge. We aim to enable systematic design of large RSNNs via a new scalable RSNN architecture and automated architectural optimization. We compose RSNNs based on a layer architecture called Sparsely-Connected Recurrent Motif Layer (SC-ML) that consists of multiple small recurrent motifs wired together by sparse lateral connections. The small size of the motifs and sparse inter-motif connectivity leads to an RSNN architecture scalable to large network sizes. We further propose a method called Hybrid Risk-Mitigating Architectural Search (HRMAS) to systematically optimize the topology of the proposed recurrent motifs and SC-ML layer architecture. HRMAS is an alternating two-step optimization process by which we mitigate the risk of network instability and performance degradation caused by architectural change by introducing a novel biologically-inspired "self-repairing" mechanism through intrinsic plasticity. The intrinsic plasticity is introduced to the second step of each HRMAS iteration and acts as unsupervised fast self-adaptation to structural and synaptic weight modifications introduced by the first step during the RSNN architectural "evolution." We demonstrate that the proposed automatic architecture optimization leads to significant performance gains over existing manually designed RSNNs: we achieve 96.44% on TI46-Alpha, 94.66% on N-TIDIGITS, 90.28% on DVS-Gesture, and 98.72% on N-MNIST. To the best of the authors' knowledge, this is the first work to perform systematic architecture optimization on RSNNs.

NoC simulation steered by NEST: McAERsim and a Noxim patch.

Introduction: Great knowledge was gained about the computational substrate of the brain, but the way in which components and entities interact to perform information processing still remains a secret. Complex and large-scale network models have been developed to unveil processes at the ensemble level taking place over a large range of timescales. They challenge any kind of simulation platform, so that efficient implementations need to be developed that gain from focusing on a set of relevant models. With increasing network sizes imposed by these models, low latency inter-node communication becomes a critical aspect. This situation is even accentuated, if slow processes like learning should be covered, that require faster than real-time simulation. Methods: Therefore, this article presents two simulation frameworks, in which network-on-chip simulators are interfaced with the neuroscientific development environment NEST. This combination yields network traffic that is directly defined by the relevant neural network models and used to steer the network-on-chip simulations. As one of the outcomes, instructive statistics on network latencies are obtained. Since time stamps of different granularity are used by the simulators, a conversion is required that can be exploited to emulate an intended acceleration factor. Results: By application of the frameworks to scaled versions of the cortical microcircuit model-selected because of its unique properties as well as challenging demands-performance curves, latency, and traffic distributions could be determined. Discussion: The distinct characteristic of the second framework is its tree-based source-address driven multicast support, which, in connection with the torus topology, always led to the best results. Although currently biased by some inherent assumptions of the network-on-chip simulators, the results suit well to those of previous work dealing with node internals and suggesting accelerated simulations to be in reach.

Incorporating slow NMDA-type receptors with nonlinear voltage-dependent magnesium block in a next generation neural mass model: derivation and dynamics.

We derive a next generation neural mass model of a population of quadratic-integrate-and-fire neurons, with slow adaptation, and conductance-based AMPAR, GABAR and nonlinear NMDAR synapses. We show that the Lorentzian ansatz assumption can be satisfied by introducing a piece-wise polynomial approximation of the nonlinear voltage-dependent magnesium block of NMDAR current. We study the dynamics of the resulting system for two example cases of excitatory cortical neurons and inhibitory striatal neurons. Bifurcation diagrams are presented comparing the different dynamical regimes as compared to the case of linear NMDAR currents, along with sample comparison simulation time series demonstrating different possible oscillatory solutions. The omission of the nonlinearity of NMDAR currents results in a shift in the range (and possible disappearance) of the constant high firing rate regime, along with a modulation in the amplitude and frequency power spectrum of oscillations. Moreover, nonlinear NMDAR action is seen to be state-dependent and can have opposite effects depending on the type of neurons involved and the level of input firing rate received. The presented model can serve as a computationally efficient building block in whole brain network models for investigating the differential modulation of different types of synapses under neuromodulatory influence or receptor specific malfunction.

Spike-Based Neuromorphic Hardware for Dynamic Tactile Perception with a Self-Powered Mechanoreceptor Array.

A self-powered mechanoreceptor array is demonstrated using four mechanoreceptor cells for recognition of dynamic touch gestures. Each cell consists of a triboelectric nanogenerator (TENG) for touch sensing and a bi-stable resistor (biristor) for spike encoding. It produces informative spike signals by sensing a force of an external touch and encoding the force into the number of spikes. An array of the mechanoreceptor cells is utilized to monitor various touch gestures and it successfully generated spike signals corresponding to all the gestures. To validate the practicality of the mechanoreceptor array, a spiking neural network (SNN), highly attractive for power consumption compared to the conventional von Neumann architecture, is used for the identification of touch gestures. The measured spiking signals are reflected as inputs for the SNN simulations. Consequently, touch gestures are classified with a high accuracy rate of 92.5%. The proposed mechanoreceptor array emerges as a promising candidate for a building block of tactile in-sensor computing in the era of the Internet of Things (IoT), due to the low cost and high manufacturability of the TENG. This eliminates the need for a power supply, coupled with the intrinsic high throughput of the Si-based biristor employing complementary metal-oxide-semiconductor (CMOS) technology.

Modeling the Interictal Epileptic State for Therapeutic Development with Tetanus Toxin.

Focal forms of epilepsy can result from a wide range of insults and can vary from focal symptoms to generalized convulsions. Most drugs that have been developed for epilepsy focus on the prevention of seizures. On Electroencephalography (EEG), seizures are characterized by a repetitive buildup of epileptic waveforms that can spread across the brain. Brain regions that produce seizures generate far more frequent 'interictal' spikes seen between seizures, and in animal models, these spikes occur prior to the development of seizures. Interictal spiking by itself has been shown to have significant adverse clinical effects on cognition and behavior in both patients and animal models. While the exact relationships between interictal spiking and seizures are not well defined, interictal spikes serve as an important biomarker that, for some forms of epilepsy, can serve as a surrogate biomarker and as a druggable target. While there are many animal models of seizures for drug development, here we review models of interictal spiking, focusing on tetanus toxin, to study the relationship between interictal spiking, seizures, cognition, and behavior. Studies on human cortical regions with frequent interictal spiking have identified potential therapeutic targets; therefore, having a highly consistent model of spiking will be invaluable not only for unraveling the initial stages of the pathological cascade leading to seizure development but also for testing novel therapeutics. This review offers a succinct overview of the use of tetanus toxin animal models for studying and therapeutic development for interictal spiking.

Spiking Laguerre Volterra networks-predicting neuronal activity from local field potentials.

Objective.Understanding the generative mechanism between local field potentials (LFP) and neuronal spiking activity is a crucial step for understanding information processing in the brain. Up to now, most approaches have relied on simply quantifying the coupling between LFP and spikes. However, very few have managed to predict the exact timing of spike occurrence based on LFP variations.Approach.Here, we fill this gap by proposing novel spiking Laguerre-Volterra network (sLVN) models to describe the dynamic LFP-spike relationship. Compared to conventional artificial neural networks, the sLVNs are interpretable models that provide explainable features of the underlying dynamics.Main results.The proposed networks were applied on extracellular microelectrode recordings of Parkinson's Disease patients during deep brain stimulation (DBS) surgery. Based on the predictability of the LFP-spike pairs, we detected three neuronal populations with unique signal characteristics and sLVN model features.Significance.These clusters were indirectly associated with motor score improvement following DBS surgery, warranting further investigation into the potential of spiking activity predictability as an intraoperative biomarker for optimal DBS lead placement.

LDD: High-Precision Training of Deep Spiking Neural Network Transformers Guided by an Artificial Neural Network.

The rise of large-scale Transformers has led to challenges regarding computational costs and energy consumption. In this context, spiking neural networks (SNNs) offer potential solutions due to their energy efficiency and processing speed. However, the inaccuracy of surrogate gradients and feature space quantization pose challenges for directly training deep SNN Transformers. To tackle these challenges, we propose a method (called LDD) to align ANN and SNN features across different abstraction levels in a Transformer network. LDD incorporates structured feature knowledge from ANNs to guide SNN training, ensuring the preservation of crucial information and addressing inaccuracies in surrogate gradients through designing layer-wise distillation losses. The proposed approach outperforms existing methods on the CIFAR10 (96.1%), CIFAR100 (82.3%), and ImageNet (80.9%) datasets, and enables training of the deepest SNN Transformer network using ImageNet.

A nanofluidic spiking synapse.

Currently, the nanofluidic synapse can only perform basic neuromorphic pulse patterns. One immediate problem that needs to be addressed to further its capability of brain-like computing is the realization of a nanofluidic spiking device. Here, we report the use of a poly(3,4-ethylenedioxythiophene) polystyrene sulfonate membrane to achieve bionic ionic current-induced spiking. In addition to the simulation of various electrical pulse patterns, our synapse could produce transmembrane ionic current-induced spiking, which is highly analogous to biological action potentials with similar phases and excitability. Moreover, the spiking properties could be modulated by ions and neurochemicals. We expect that this work could contribute to biomimetic spiking computing in solution.

A lightweight data-driven spiking neuronal network model of Drosophila olfactory nervous system with dedicated hardware support.

Data-driven spiking neuronal network (SNN) models enable in-silico analysis of the nervous system at the cellular and synaptic level. Therefore, they are a key tool for elucidating the information processing principles of the brain. While extensive research has focused on developing data-driven SNN models for mammalian brains, their complexity poses challenges in achieving precision. Network topology often relies on statistical inference, and the functions of specific brain regions and supporting neuronal activities remain unclear. Additionally, these models demand huge computing facilities and their simulation speed is considerably slower than real-time. Here, we propose a lightweight data-driven SNN model that strikes a balance between simplicity and reproducibility. The model is built using a qualitative modeling approach that can reproduce key dynamics of neuronal activity. We target the Drosophila olfactory nervous system, extracting its network topology from connectome data. The model was successfully implemented on a small entry-level field-programmable gate array and simulated the activity of a network in real-time. In addition, the model reproduced olfactory associative learning, the primary function of the olfactory system, and characteristic spiking activities of different neuron types. In sum, this paper propose a method for building data-driven SNN models from biological data. Our approach reproduces the function and neuronal activities of the nervous system and is lightweight, acceleratable with dedicated hardware, making it scalable to large-scale networks. Therefore, our approach is expected to play an important role in elucidating the brain's information processing at the cellular and synaptic level through an analysis-by-construction approach. In addition, it may be applicable to edge artificial intelligence systems in the future.

Similar neural states, but dissimilar decoding patterns for motor control in parietal cortex.

Discrete neural states are associated with reaching movements across the fronto-parietal network. Here, the Hidden Markov Model (HMM) applied to spiking activity of the somato-motor parietal area PE revealed a sequence of states similar to those of the contiguous visuomotor areas PEc and V6A. Using a coupled clustering and decoding approach, we proved that these neural states carried spatiotemporal information regarding behaviour in all three posterior parietal areas. However, comparing decoding accuracy, PE was less informative than V6A and PEc. In addition, V6A outperformed PEc in target inference, indicating functional differences among the parietal areas. To check the consistency of these differences, we used both a supervised and an unsupervised variant of the HMM, and compared its performance with two more common classifiers, Support Vector Machine and Long-Short Term Memory. The differences in decoding between areas were invariant to the algorithm used, still showing the dissimilarities found with HMM, thus indicating that these dissimilarities are intrinsic in the information encoded by parietal neurons. These results highlight that, when decoding from the parietal cortex, for example, in brain machine interface implementations, attention should be paid in selecting the most suitable source of neural signals, given the great heterogeneity of this cortical sector.

Applying HMMs to spiking activity recorded from the somato-motor parietal area PE revealed discrete neural states related to reaching movements. These states were extremely similar to those present in the neighbouring visuomotor areas PEc and V6A. Our decoding approach showed that these states conveyed spatiotemporal behaviour information across all three posterior parietal areas. However, decoding accuracy was lower in PE compared to V6A and PEc, with V6A excelling in target inference. These differences held true even when changing the decoding algorithm, indicating intrinsic dissimilarities in information encoding by parietal different areas. These findings highlight the importance of selecting the appropriate neural signal sources in applications such as brain machine interfaces and pave the way for further investigation of the nontrivial diversity within the parietal cortex.

Directly training temporal Spiking Neural Network with sparse surrogate gradient.

Brain-inspired Spiking Neural Networks (SNNs) have attracted much attention due to their event-based computing and energy-efficient features. However, the spiking all-or-none nature has prevented direct training of SNNs for various applications. The surrogate gradient (SG) algorithm has recently enabled spiking neural networks to shine in neuromorphic hardware. However, introducing surrogate gradients has caused SNNs to lose their original sparsity, thus leading to the potential performance loss. In this paper, we first analyze the current problem of direct training using SGs and then propose Masked Surrogate Gradients (MSGs) to balance the effectiveness of training and the sparseness of the gradient, thereby improving the generalization ability of SNNs. Moreover, we introduce a temporally weighted output (TWO) method to decode the network output, reinforcing the importance of correct timesteps. Extensive experiments on diverse network structures and datasets show that training with MSG and TWO surpasses the SOTA technique.

A Delayed Spiking Neural Membrane System for Adaptive Nearest Neighbor-Based Density Peak Clustering.

Although the density peak clustering (DPC) algorithm can effectively distribute samples and quickly identify noise points, it lacks adaptability and cannot consider the local data structure. In addition, clustering algorithms generally suffer from high time complexity. Prior research suggests that clustering algorithms grounded in P systems can mitigate time complexity concerns. Within the realm of membrane systems (P systems), spiking neural P systems (SN P systems), inspired by biological nervous systems, are third-generation neural networks that possess intricate structures and offer substantial parallelism advantages. Thus, this study first improved the DPC by introducing the maximum nearest neighbor distance and K-nearest neighbors (KNN). Moreover, a method based on delayed spiking neural P systems (DSN P systems) was proposed to improve the performance of the algorithm. Subsequently, the DSNP-ANDPC algorithm was proposed. The effectiveness of DSNP-ANDPC was evaluated through comprehensive evaluations across four synthetic datasets and 10 real-world datasets. The proposed method outperformed the other comparison methods in most cases.

Seizure Detection of EEG Signals Based on Multi-Channel Long- and Short-Term Memory-Like Spiking Neural Model.

Seizure is a common neurological disorder that usually manifests itself in recurring seizure, and these seizures can have a serious impact on a person's life and health. Therefore, early detection and diagnosis of seizure is crucial. In order to improve the efficiency of early detection and diagnosis of seizure, this paper proposes a new seizure detection method, which is based on discrete wavelet transform (DWT) and multi-channel long- and short-term memory-like spiking neural P (LSTM-SNP) model. First, the signal is decomposed into 5 levels by using DWT transform to obtain the features of the components at different frequencies, and a series of time-frequency features in wavelet coefficients are extracted. Then, these different features are used to train a multi-channel LSTM-SNP model and perform seizure detection. The proposed method achieves a high seizure detection accuracy on the CHB-MIT dataset: 98.25% accuracy, 98.22% specificity and 97.59% sensitivity. This indicates that the proposed epilepsy detection method can show competitive detection performance.

Reliable object tracking by multimodal hybrid feature extraction and transformer-based fusion.

Visual object tracking, which is primarily based on visible light image sequences, encounters numerous challenges in complicated scenarios, such as low light conditions, high dynamic ranges, and background clutter. To address these challenges, incorporating the advantages of multiple visual modalities is a promising solution for achieving reliable object tracking. However, the existing approaches usually integrate multimodal inputs through adaptive local feature interactions, which cannot leverage the full potential of visual cues, thus resulting in insufficient feature modeling. In this study, we propose a novel multimodal hybrid tracker (MMHT) that utilizes frame-event-based data for reliable single object tracking. The MMHT model employs a hybrid backbone consisting of an artificial neural network (ANN) and a spiking neural network (SNN) to extract dominant features from different visual modalities and then uses a unified encoder to align the features across different domains. Moreover, we propose an enhanced transformer-based module to fuse multimodal features using attention mechanisms. With these methods, the MMHT model can effectively construct a multiscale and multidimensional visual feature space and achieve discriminative feature modeling. Extensive experiments demonstrate that the MMHT model exhibits competitive performance in comparison with that of other state-of-the-art methods. Overall, our results highlight the effectiveness of the MMHT model in terms of addressing the challenges faced in visual object tracking tasks.