探索深度神经网络中结构化剪枝与容错性之间的权衡：面向空间应用
Exploring the Trade-Off Between Structured Pruning and Fault Tolerance in Deep Neural Networks for Space Applications
浏览论文内容
中文总结 AI 辅助
本文实证研究深度神经网络宽度与容错性的权衡，发现结构化剪枝虽增加单次推理故障敏感性，但缩短执行时间可抵消风险，实现节能降延迟且不损可靠性，为空间AI系统设计提供指导。
中文摘要 AI 辅助
深度神经网络（DNNs）由于其信息的分布式表示，固有地表现出一定程度的对位级故障的鲁棒性。随着模型宽度的增加，信息变得更加分散，理论上降低了任何单个位故障的影响。在本文中，我们实证研究了模型宽度与对单事件翻转（SEUs）鲁棒性之间的关系。我们进行了一项全面的实验，其中基线模型经过迭代的结构化剪枝以减小宽度，同时尽可能保持任务性能。在每个剪枝阶段，我们运行一次有针对性的故障注入活动，以评估模型在模拟位翻转场景下的性能。我们的结果表明，尽管结构化剪枝通过减少冗余增加了每次推理对故障的敏感性，但这一效应被更短的执行时间有效抵消，后者降低了遇到SEU的概率。这些发现表明，结构化剪枝可以在不损害整体可靠性的情况下带来显著的能源和延迟节省，为设计面向空间应用的鲁棒AI系统提供了有用的指导。
英文摘要
Deep Neural Networks (DNNs) inherently exhibit a degree of robustness to bit-level faults due to their distributed representation of information. As a model increases in width, this information becomes more dispersed, theoretically reducing the impact of any single bit fault. In this paper, we empirically investigate the relationship between model width and robustness to Single Event Upsets (SEUs). We conduct a comprehensive experiment in which baseline models undergo iterative structured pruning to reduce their width while preserving task performance as much as possible. At each pruning stage, we run a targeted fault-injection campaign to evaluate the model's performance under simulated bit-flip scenarios. Our results show that, although structured pruning increases per-inference sensitivity to faults by reducing redundancy, this effect is effectively counterbalanced by shorter execution time, which lowers the probability of encountering an SEU. These findings suggest that structured pruning can yield significant energy and latency savings without compromising overall reliability, providing useful guidance for designing robust AI systems for space applications.
发表机构
- Magics Technologies
- KU Leuven（鲁汶大学）
- Flanders Make
机构由 AI 辅助整理，请以论文原文为准。
