SPEAR: Structured Pruning for Spiking Neural Networks via Synaptic Operation Estimation and Reinforcement Learning

2025-06-28

AI Summary

Deploying spiking neural networks (SNNs) on resource-constrained neuromorphic hardware remains challenging due to difficulty in satisfying strict synaptic operation (SynOps) constraints during pruning. Method: This paper proposes a structured pruning framework that, for the first time, treats SynOps as a hard constraint in reinforcement learning (RL)-based architecture search. It introduces a lightweight SynOps estimator (LRE) for real-time post-pruning SynOps prediction and a task-adaptive reward (TAR) function that dynamically balances accuracy degradation against constraint satisfaction, thereby preventing constraint violations during search. Contribution/Results: Under stringent SynOps budgets, the framework achieves significant SNN model compression and reduced computational cost while preserving high accuracy. Extensive experiments across multiple benchmark tasks demonstrate its effectiveness for efficient edge deployment on neuromorphic hardware.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

While deep spiking neural networks (SNNs) demonstrate superior performance, their deployment on resource-constrained neuromorphic hardware still remains challenging. Network pruning offers a viable solution by reducing both parameters and synaptic operations (SynOps) to facilitate the edge deployment of SNNs, among which search-based pruning methods search for the SNNs structure after pruning. However, existing search-based methods fail to directly use SynOps as the constraint because it will dynamically change in the searching process, resulting in the final searched network violating the expected SynOps target. In this paper, we introduce a novel SNN pruning framework called SPEAR, which leverages reinforcement learning (RL) technique to directly use SynOps as the searching constraint. To avoid the violation of SynOps requirements, we first propose a SynOps prediction mechanism called LRE to accurately predict the final SynOps after search. Observing SynOps cannot be explicitly calculated and added to constrain the action in RL, we propose a novel reward called TAR to stabilize the searching. Extensive experiments show that our SPEAR framework can effectively compress SNN under specific SynOps constraint.

Problem

• Reducing SNN parameters and SynOps for edge deployment
• Accurately predicting SynOps to meet target constraints
• Stabilizing RL-based search with novel reward mechanism

Innovation

• Uses reinforcement learning for SNN pruning
• Predicts SynOps with LRE mechanism
• Stabilizes search with TAR reward