Write a Blog >>
Tue 10 Nov 2020 01:43 - 01:44 at Virtual room 1 - ML Testing 1

Trained DNN models are increasingly adopted as integral parts of software systems, but they often perform deficiently in the field. A particularly damaging problem is that DNN models often give false predictions with high confidence, due to the unavoidable slight divergences between operation data and training data. To minimize the loss caused by inaccurate confidence, operational calibration, i.e., calibrating the confidence function of a DNN classifier against its operation domain, becomes a necessary debugging step in the engineering of the whole system.

Operational calibration is difficult considering the limited budget of labeling operation data and the weak interpretability of DNN models. We propose a Bayesian approach to operational calibration that gradually corrects the confidence given by the model under calibration with a small number of labeled operation data deliberately selected from a larger set of unlabeled operation data. The approach is made effective and efficient by leveraging the locality of the learned representation of the DNN model and modeling the calibration as Gaussian Process Regression. Comprehensive experiments with various practical datasets and DNN models show that it significantly outperformed alternative methods, and in some difficult tasks it eliminated about 71% to 97% high-confidence (>0.9) errors with only about 10% of the minimal amount of labeled operation data needed for practical learning techniques to barely work

Tue 10 Nov

Displayed time zone: (UTC) Coordinated Universal Time change

01:30 - 02:00
01:30
2m
Talk
Correlations between Deep Neural Network Model Coverage Criteria and Model Quality
Research Papers
Shenao Yan Rutgers University, USA, Guanhong Tao Purdue University, USA, Xuwei Liu Purdue University, USA, Juan Zhai Rutgers University, USA, Shiqing Ma Rutgers University, USA, Lei Xu Nanjing University, China, Xiangyu Zhang Purdue University
DOI
01:33
1m
Talk
Deep Learning Library Testing via Effective Model GenerationACM SIGSOFT Distinguished Paper Award
Research Papers
Zan Wang Tianjin University, China, Ming Yan Tianjin University, China, Junjie Chen Tianjin University, China, Shuang Liu Tianjin University, China, Dongdi Zhang Tianjin University, China
DOI
01:35
1m
Talk
Detecting Numerical Bugs in Neural Network ArchitecturesACM SIGSOFT Distinguished Paper Award
Research Papers
Yuhao Zhang Peking University, Luyao Ren Peking University, China, Liqian Chen National University of Defense Technology, China, Yingfei Xiong Peking University, Shing-Chi Cheung Hong Kong University of Science and Technology, China, Tao Xie Peking University
DOI
01:37
1m
Talk
Dynamic Slicing for Deep Neural Networks
Research Papers
Ziqi Zhang Peking University, China, Yuanchun Li Microsoft Research, China, Yao Guo Peking University, Xiangqun Chen Peking University, Yunxin Liu Microsoft Research, China
DOI
01:39
1m
Talk
Grammar Based Directed Testing of Machine Learning Systems
Journal First
Sakshi Udeshi Singapore University of Technology and Design, Sudipta Chattopadhyay Singapore University of Technology and Design
01:41
1m
Talk
Is Neuron Coverage a Meaningful Measure for Testing Deep Neural Networks?
Research Papers
Fabrice Harel-Canada University of California at Los Angeles, USA, Lingxiao Wang University of California at Los Angeles, USA, Muhammad Ali Gulzar University of California at Los Angeles, USA, Quanquan Gu University of California at Los Angeles, USA, Miryung Kim University of California at Los Angeles, USA
DOI
01:43
1m
Talk
Operational Calibration: Debugging Confidence Errors for DNNs in the Field
Research Papers
Zenan Li Nanjing University, China, Xiaoxing Ma Nanjing University, China, Chang Xu Nanjing University, China, Jingwei Xu Nanjing University, China, Chun Cao Nanjing University, China, Jian Lu Nanjing University, China
DOI
01:45
15m
Talk
Conversations on ML Testing 1
Research Papers
Fabrice Harel-Canada University of California at Los Angeles, USA, Ming Yan Tianjin University, China, Sakshi Udeshi Singapore University of Technology and Design, Shenao Yan Rutgers University, USA, Yuhao Zhang Peking University, Zenan Li Nanjing University, China, Ziqi Zhang Peking University, China, M: Hamid Bagheri University of Nebraska-Lincoln, USA