Faster and Stronger: When ANN-SNN Conversion Meets Parallel Spiking Calculation

2024-12-18

arXiv.org

AI Summary

Spiking Neural Networks (SNNs) face fundamental bottlenecks: high training overhead, excessive inference latency, and poor scalability to complex tasks. To address these, we propose the first learning framework integrating parallel spike computation with ANN-to-SNN conversion. Our method establishes a rigorous mathematical mapping between per-timestep spike rates and cumulative spike counts, enabling training-free, activation-function-agnostic conversion. We theoretically prove that this conversion is lossless and order-preserving. Furthermore, we introduce an optimal time-shift distance estimator and a distribution-aware error calibration mechanism to minimize conversion-induced approximation errors. Evaluated across diverse architectures and vision/speech benchmarks, our approach achieves state-of-the-art accuracy under ultra-low latency—requiring only 1–4 timesteps—significantly outperforming STBP and mainstream conversion methods.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Natural Language Processing: Learning & Optimization for NLP
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Social networks and social learning

Abstract

Spiking Neural Network (SNN), as a brain-inspired and energy-efficient network, is currently facing the pivotal challenge of exploring a suitable and efficient learning framework. The predominant training methodologies, namely Spatial-Temporal Back-propagation (STBP) and ANN-SNN Conversion, are encumbered by substantial training overhead or pronounced inference latency, which impedes the advancement of SNNs in scaling to larger networks and navigating intricate application domains. In this work, we propose a novel parallel conversion learning framework, which establishes a mathematical mapping relationship between each time-step of the parallel spiking neurons and the cumulative spike firing rate. We theoretically validate the lossless and sorting properties of the conversion process, as well as pointing out the optimal shifting distance for each step. Furthermore, by integrating the above framework with the distribution-aware error calibration technique, we can achieve efficient conversion towards more general activation functions or training-free circumstance. Extensive experiments have confirmed the significant performance advantages of our method for various conversion cases under ultra-low time latency. To our best knowledge, this is the first work which jointly utilizes parallel spiking calculation and ANN-SNN Conversion, providing a highly promising approach for SNN supervised training.

Problem

• Overcoming high training overhead in SNN learning frameworks
• Reducing inference latency in ANN-SNN conversion methods
• Enabling efficient SNN scaling for complex applications

Innovation

• Parallel conversion learning framework for SNNs
• Mathematical mapping for spike firing rates
• Distribution-aware error calibration technique