Problem
Research questions and friction points this paper is trying to address.
Structured Pruning
Fault Tolerance
Single Event Upsets
Deep Neural Networks
Space Applications
Exploring the Trade-Off Between Structured Pruning and Fault Tolerance in Deep Neural Networks for Space Applications
🤖 AI Summary
This study investigates the trade-off between structured pruning of deep neural networks and fault tolerance in space applications. Addressing the threat of single-event upsets in aerospace environments, the authors iteratively prune model width while injecting faults to systematically evaluate the balance between robustness and execution efficiency. The findings reveal that although pruning increases parameter sensitivity, this adverse effect is effectively offset by the reduced execution time. Results demonstrate that structured pruning significantly lowers energy consumption and latency while maintaining overall reliability. This work provides critical guidance for designing energy-efficient and highly reliable AI systems for aerospace missions.
📝 Abstract
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.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Structured Pruning
Fault Tolerance
Single Event Upsets
Deep Neural Networks
Space Applications
🔎 Similar Papers
No similar papers found.
