spiking neural modeling

This work addresses the inefficiencies of conventional neural networks in data and energy consumption, as well as the limited availability of effective learning algorithms for spiking neural networks (SNNs). To bridge this gap, the authors propose Spark, a modular SNN framework that constructs end-to-end models by composing simple plasticity-based components, enabling continuous, batch-free learning. Spark integrates seamlessly with traditional machine learning pipelines while incorporating biologically inspired continual learning mechanisms, substantially improving data efficiency and practical applicability. The framework’s effectiveness is demonstrated on the sparse-reward CartPole task, where it successfully learns in a continuous setting, highlighting the promise of modular design for efficient SNN training.

This work addresses the fundamental challenge in spiking neural networks (SNNs) of reconciling convolutional weight sharing with the biological constraint of local synaptic plasticity. To this end, we propose a biologically inspired convolutional SNN architecture for image classification. Our method comprises two key components: (1) employing fixed, pre-defined convolutional kernels for feedforward feature extraction—thereby preserving locality in spike-timing-dependent plasticity (STDP) or other local learning rules—and (2) introducing a data-driven initialization strategy based on domain-specific image sets to enhance both the biological plausibility and discriminative power of the initial kernels. Evaluated on the NEOVISION2 benchmark, our approach significantly improves feature representation capability and inference efficiency while maintaining high classification accuracy and strong biological interpretability. To the best of our knowledge, this is the first demonstration that a fixed convolutional structure can be effectively co-optimized with spiking dynamics to achieve competitive performance, thereby validating a novel paradigm for biologically grounded deep SNN design.

To address the challenge of simultaneously achieving biological plausibility, hardware efficiency, and competitive performance in spiking neural networks (SNNs) for general supervised classification, this paper proposes a columnar hierarchical SNN architecture tailored for classification. It employs intra-class-difference-driven columnar organization—each column represents a discriminative subcategory—and adopts an all-spiking signal flow with functionally specialized neurons. A biologically grounded learning mechanism is introduced, integrating local anti-Hebbian plasticity with dopamine neuromodulation to replace backpropagation entirely. The method unifies model-driven reinforcement learning with a state-proximity evaluation framework, enabling end-to-end training directly in the spike domain. Experiments demonstrate that the architecture achieves high accuracy and strong generalization across multiple benchmark classification tasks, while exhibiting exceptional compatibility with low-power neuromorphic hardware. This work establishes a novel paradigm for practical, deployable SNNs.

Existing ANN-to-SNN conversion methods primarily target CNNs and struggle to accommodate the nonlinear components inherent in Transformers. To address this, this paper introduces the first efficient SNN conversion framework specifically designed for Transformer architectures. Our approach features two core innovations: (1) an Expectation Compensation Module (ECM) that corrects spike-generation bias to preserve accuracy; and (2) multi-threshold spiking neurons coupled with a parallel parameter normalization mechanism, drastically reducing required timesteps and energy consumption. Evaluated on ImageNet, our converted SNN achieves 88.60% Top-1 accuracy within only four timesteps—merely 1% lower than the source ANN—while reducing power consumption by 65% (i.e., to 35% of the ANN’s). This yields a unified solution achieving high accuracy, ultra-low latency, and significant energy efficiency. To the best of our knowledge, this work establishes the first scalable, high-performance paradigm for deploying large-scale spiking neural networks based on Transformer architectures.

How do diverse neuronal temporal parameters—such as conduction delays, membrane time constants, and bursting dynamics—functionally support efficient and robust temporal information processing? Method: We employ spiking neural networks (SNNs) trained under a parameter-freezing paradigm, evaluated on controllable temporal-complexity tasks, and optimized via biologically inspired dynamic parameter adaptation. Contribution/Results: We demonstrate for the first time that tuning only neuronal temporal parameters suffices to solve multiple benchmark temporal tasks. Introducing plastic bursting mechanisms improves accuracy on complex spatiotemporal tasks by over 35%. Furthermore, synergistic spatiotemporal parameter co-adaptation markedly enhances noise robustness. These findings uncover an evolutionary and computational principle: biological neuronal diversity enables resource-efficient, resilient temporal computation under physiological constraints—revealing a fundamental mechanism underlying neural temporal coding.

Traditional artificial neural networks (ANNs) suffer from low energy efficiency and slow dynamic response, limiting their suitability for edge and neuromorphic computing. Method: This study systematically investigates modeling and training mechanisms of leaky integrate-and-fire (LIF) spiking neural networks (SNNs) for brain-inspired computing. We propose a multi-strategy comparative framework integrating surrogate gradient descent, ANN-to-SNN conversion, and spike-timing-dependent plasticity (STDP), evaluated across accuracy, energy consumption (mJ/inference), latency (ms), and convergence epochs. Results: Surrogate gradient training achieves near-ANN accuracy (within 1–2% degradation), converges by epoch 20, and attains inference latency as low as 10 ms; STDP yields ultra-low energy consumption (5 mJ/inference), enabling unsupervised learning and ultra-low-power edge deployment. This work quantitatively characterizes, for the first time, the fundamental trade-offs among accuracy, energy efficiency, and latency across SNN training paradigms—providing a deployable optimization roadmap for robotic perception, neuromorphic vision, and edge AI.

Existing theoretical analyses of spiking neural networks (SNNs) lack quantitative characterizations of their universal representational capacity as sequence-to-sequence processors over spike trains. Method: We propose a function approximation framework grounded in spike-train modeling, integrating constructive weight design with rigorous temporal complexity quantification. Contribution/Results: We establish the first constructively provable, near-optimal universal approximation theorem for naturally spikable function classes—i.e., functions admitting efficient spike-based realization. Theorematically, SNNs achieve near-optimal complexity in both neuron count and synaptic weight count. They exhibit significant representational advantages for sparse inputs, low-order temporal functions, and composite functions. Moreover, our analysis provides rigorous theoretical foundations for modular deep SNN architectures and downstream tasks such as spike-sequence classification.

This work addresses the limitations of existing spiking neurons—namely insufficient performance, poor adaptability, and low training efficiency—in large-scale vision and language tasks. The authors propose Adaptive Spiking Neurons (ASN) and their normalized variant (NASN), the first neuron designs systematically guided by a functional perspective. By incorporating learnable membrane potential dynamics, an integer-based training–spiking inference paradigm, and normalization mechanisms, ASN significantly enhances training stability and model generalizability. Comprehensive experiments across five task categories and nineteen datasets spanning both vision and language domains demonstrate that the ASN family exhibits strong generalization capabilities and holds substantial promise as a universal spiking neuron model.

This work addresses the long-standing challenges in spiking neural network (SNN) training research—namely, the lack of a systematic taxonomy and unified evaluation protocols, which have led to poor reproducibility and fragmented progress. To this end, we propose the first fine-grained, systematic classification framework for SNN training algorithms and introduce NeuroTrain, an open-source benchmark built upon snnTorch. NeuroTrain enables, for the first time, modular integration and fair comparison across diverse algorithmic paradigms, including surrogate gradient backpropagation, local and three-factor learning rules, biologically plausible plasticity mechanisms, ANN-to-SNN conversion methods, and unconventional optimization strategies. The framework standardizes the implementation of representative algorithms and supports consistent evaluation across datasets, architectures, and training configurations, thereby significantly enhancing the reproducibility and systematic investigation of SNN training methodologies.

Spiking neural networks (SNNs) suffer from limited performance on visual tasks, primarily due to inadequate information representation in conventional spike encoding schemes and poor compatibility with surrogate gradient-based training. To address this, we propose the first hybrid time-bit spike encoding framework designed explicitly for end-to-end surrogate gradient training. Our method decomposes input images into bit planes and applies temporal coding—such as latency or phase coding—to each bit plane, yielding a joint rate-time representation. Crucially, the entire encoding process is differentiable, enabling stable and efficient backpropagation via surrogate gradients. This design substantially enhances spatiotemporal information capacity and gradient propagation stability. Extensive experiments on CIFAR-10/100 and ImageNet-1K demonstrate that SNNs trained with our framework match or surpass state-of-the-art encoding methods—including rate coding, time-to-first-spike (TTFS), and spike-time coding—in accuracy, validating its effectiveness, generalizability, and seamless integration with standard surrogate-gradient optimization pipelines.