Improved Detection and Diagnosis of Faults in Deep Neural Networks Using Hierarchical and Explainable Classification
SE for AI
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.
Fri 2 MayDisplayed time zone: Eastern Time (US & Canada) change
16:00 - 17:30 | SE for AI with Quality 3Research Track / SE In Practice (SEIP) at 215 Chair(s): Sumon Biswas Case Western Reserve University | ||
16:00 15mTalk | Improved Detection and Diagnosis of Faults in Deep Neural Networks Using Hierarchical and Explainable ClassificationSE for AI Research Track Sigma Jahan Dalhousie University, Mehil Shah Dalhousie University, Parvez Mahbub Dalhousie University, Masud Rahman Dalhousie University Pre-print | ||
16:15 15mTalk | Lightweight Concolic Testing via Path-Condition Synthesis for Deep Learning LibrariesSE for AI Research Track Sehoon Kim , Yonghyeon Kim UNIST, Dahyeon Park UNIST, Yuseok Jeon UNIST, Jooyong Yi UNIST, Mijung Kim UNIST | ||
16:30 15mTalk | Mock Deep Testing: Toward Separate Development of Data and Models for Deep LearningSE for AI Research Track Ruchira Manke Tulane University, USA, Mohammad Wardat Oakland University, USA, Foutse Khomh Polytechnique Montréal, Hridesh Rajan Tulane University | ||
16:45 15mTalk | RUG: Turbo LLM for Rust Unit Test GenerationSE for AI Research Track Xiang Cheng Georgia Institute of Technology, Fan Sang Georgia Institute of Technology, Yizhuo Zhai Georgia Institute of Technology, Xiaokuan Zhang George Mason University, Taesoo Kim Georgia Institute of Technology Pre-print Media Attached File Attached | ||
17:00 15mTalk | Test Input Validation for Vision-based DL Systems: An Active Learning Approach SE In Practice (SEIP) Delaram Ghobari University of Ottawa, Mohammad Hossein Amini University of Ottawa, Dai Quoc Tran SmartInsideAI Company Ltd. and Sungkyunkwan University, Seunghee Park SmartInsideAI Company Ltd. and Sungkyunkwan University, Shiva Nejati University of Ottawa, Mehrdad Sabetzadeh University of Ottawa Pre-print | ||
17:15 15mTalk | SEMANTIC CODE FINDER: An Efficient Semantic Search Framework for Large-Scale Codebases SE In Practice (SEIP) daeha ryu Innovation Center, Samsung Electronics, Seokjun Ko Samsung Electronics Co., Eunbi Jang Innovation Center, Samsung Electronics, jinyoung park Innovation Center, Samsung Electronics, myunggwan kim Innovation Center, Samsung Electronics, changseo park Innovation Center, Samsung Electronics |