Dynamic Sensitivity Filter Pruning using Multi-Agent Reinforcement Learning For DCNN's

2025-09-05

AI Summary

To address the computational and memory overhead bottlenecks of deploying deep convolutional neural networks (CNNs) on edge devices, this paper proposes a dynamic filter pruning framework based on multi-agent reinforcement learning. The method introduces a differential-aware importance evaluation mechanism that jointly incorporates gradient sensitivity, first-order Taylor expansion-based response approximation, and KL divergence of activation distributions. An exponential scaling strategy is further employed to precisely identify structurally unstable or non-critical filters. Pruning is completed in a single forward-backward pass, ensuring both efficiency and determinism. Under pruning ratios of 50%–70%, the framework reduces floating-point operations by over 80%; at 70% pruning, top-1 accuracy remains as high as 98.23%. It significantly outperforms conventional heuristic methods, achieving synergistic advances in compression ratio, accuracy retention, and generalization capability.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Learning & Optimization for CV
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Deep Convolutional Neural Networks have achieved state of the art performance across various computer vision tasks, however their practical deployment is limited by computational and memory overhead. This paper introduces Differential Sensitivity Fusion Pruning, a novel single shot filter pruning framework that focuses on evaluating the stability and redundancy of filter importance scores across multiple criteria. Differential Sensitivity Fusion Pruning computes a differential sensitivity score for each filter by fusing the discrepancies among gradient based sensitivity, first order Taylor expansion, and KL divergence of activation distributions. An exponential scaling mechanism is applied to emphasize filters with inconsistent importance across metrics, identifying candidates that are structurally unstable or less critical to the model performance. Unlike iterative or reinforcement learning based pruning strategies, Differential Sensitivity Fusion Pruning is efficient and deterministic, requiring only a single forward-backward pass for scoring and pruning. Extensive experiments across varying pruning rates between 50 to 70 percent demonstrate that Differential Sensitivity Fusion Pruning significantly reduces model complexity, achieving over 80 percent Floating point Operations Per Seconds reduction while maintaining high accuracy. For instance, at 70 percent pruning, our approach retains up to 98.23 percent of baseline accuracy, surpassing traditional heuristics in both compression and generalization. The proposed method presents an effective solution for scalable and adaptive Deep Convolutional Neural Networks compression, paving the way for efficient deployment on edge and mobile platforms.

Problem

• Reducing computational and memory overhead in DCNNs
• Evaluating filter importance stability and redundancy
• Achieving high compression while maintaining accuracy

Innovation

• Fuses gradient, Taylor, KL metrics for filter scoring
• Applies exponential scaling to identify unstable filters
• Single-pass deterministic pruning reduces FLOPs significantly