Light-Weight Neural Network Repair for Edge Computing Scenarios
The continuous evolution and progress of deep learning technology have ushered in an era where neural networks play a pivotal role in various fields.This is particularly evident in edge computing environments such as intelligent transportation systems and next-generation power grids.However,despite the widespread applications,the reliability of neural networks remains a significant bottleneck,restricting their full potential in real-world scenarios.One of the primary challenges arises in complex edge environments,where pre-trained models often suffer from performance degradation due to the inherent difficulty of covering all possible edge scenarios comprehensively.Consequently,the need for an efficient repair mechanism for deployed neural networks has become a paramount focus of research.Traditional methods of repairing neural networks typically involve the cumbersome process of retraining the entire model.This approach presents several challenges,especially in edge scenarios.Firstly,devices located in different geographical regions may encounter unique natural noise,making it a challenging task for a unified model to seamlessly adapt to all diverse environments.Secondly,the large-scale parameters of deep neural networks contribute to substantial resource consumption during training and deployment,and the potential for service interruptions during updates poses a threat to system availability.To address these challenges head-on,this paper proposes a lightweight patch-based neural network repair algorithm.The core objective of this algorithm is to augment the robustness of neural networks against natural noise and corner cases prevalent in different edge environments by introducing personalized patches.The fault localization phase is a key aspect of this algorithm,drawing inspiration from probes in program instrumentation used to detect,improve,and analyze software behavior.In the context of neural networks,the paper introduces neural network instrumentation technology.This involves strategically inserting model probes into the neural network to observe its internal patterns,facilitating the localization of faults in error samples.During the fault repair phase,customized patches,obtained through an innovative unsupervised search,are seamlessly integrated into the original neural network to rectify its output.Additionally,a fault prediction module is introduced,capable of predicting potential errors in advance and activating patches only when deemed necessary.In a series of comprehensive experiments leveraging 2 datasets,15 distinct types of noise,and 4 diverse neural network models,the proposed approach yielded remarkable results.Performance improvements ranging from 6.64％to an impressive 20.00％were observed when compared to existing repair algorithms.Furthermore,the innovative approach demonstrated its efficiency by drastically reducing the required quantity of training samples by over 90％.The highest reduction in required updated parameters reached an astounding 91.94％.This effective and lightweight patch-based approach not only addresses the reliability concerns of neural networks in edge computing environments but also sets the stage for more resilient and adaptable applications in the complex and dynamic real-world scenarios of the future.
NETL
NSTL
万方数据
