ExplainabilitySafetyTrustworthinesscs.SE
Improved Detection and Diagnosis of Faults in Deep Neural Networks Using Hierarchical and Explainable Classification
Sigma Jahan, Mehil B Shah, Parvez Mahbub, Mohammad Masudur Rahman
January 22, 2025
International Conference on Software Engineering
6 citations
Key Takeaways
-
DNNs are widely deployed but unreliable: Deep neural networks are used in critical applications like medical diagnosis, autonomous driving, and fraud detection, yet they "often suffer from reliability issues due to their inherent complexity," highlighting a significant gap between current capabilities and responsible deployment standards.
-
Existing fault detection methods are incomplete: The paper identifies that current techniques for finding problems in DNNs are "limited by the types of faults they support or the kind of information they use," meaning organizations may miss important failures in their AI systems without comprehensive diagnostic tools.
-
A new detection approach achieves strong performance: DEFault detects faults in real-world DNN programs with approximately 94% recall by combining runtime monitoring with explainable AI methods, outperforming previous approaches by 3.92%-11.54%, suggesting practical improvements are possible.
-
Root cause diagnosis remains challenging: While fault detection reaches 94% recall, diagnosing the actual cause of failures only achieves 63% recall, indicating that even with improved tools, pinpointing why DNNs fail remains difficult—an important consideration for organizations trying to fix problematic AI systems.
-
Explainability aids troubleshooting: The technique uses explainable AI methods (specifically SHAP) to identify root causes, demonstrating that making AI systems more interpretable can help practitioners understand and address failures rather than treating them as black boxes.
Abstract
Deep Neural Networks (DNN) have found numerous applications in various domains, including fraud detection, medical diagnosis, facial recognition, and autonomous driving. However, DNN-based systems often suffer from reliability issues due to their inherent complexity and the stochastic nature of their underlying models. Unfortunately, existing techniques to detect faults in DNN programs are either limited by the types of faults (e.g., hyperparameter or layer) they support or the kind of information (e.g., dynamic or static) they use. As a result, they might fall short of comprehensively detecting and diagnosing the faults. In this paper, we present DEFault (Detect and Explain Fault) -- a novel technique to detect and diagnose faults in DNN programs. It first captures dynamic (i.e., runtime) features during model training and leverages a hierarchical classification approach to detect all major fault categories from the literature. Then, it captures static features (e.g., layer types) from DNN programs and leverages explainable AI methods (e.g., SHAP) to narrow down the root cause of the fault. We train and evaluate DEFault on a large, diverse dataset of ~14.5K DNN programs and further validate our technique using a benchmark dataset of 52 real-life faulty DNN programs. Our approach achieves ~94% recall in detecting real-world faulty DNN programs and ~63% recall in diagnosing the root causes of the faults, demonstrating 3.92% - 11.54% higher performance than that of state-of-the-art techniques. Thus, DEFault has the potential to significantly improve the reliability of DNN programs by effectively detecting and diagnosing the faults.
Paper Info
arXiv ID2501.12560
PublishedJanuary 22, 2025
Citations
6
Relevance72%
VenueInternational Conference on Software Engineering
TypeJournalArticle, Conference
Categorycs.SE
Responsible AI Topics
ExplainabilitySafetyTrustworthiness
Links
Citation data from Semantic Scholar