Write a Blog >>
ICSE 2020
Wed 24 June - Thu 16 July 2020
Thu 9 Jul 2020 00:12 - 00:24 at Goguryeo - P14-Testing Chair(s): Shin Yoo

Deep neural networks (DNN) have been shown to be notoriously brittle to small perturbations in their input data. This problem is analogous to the over-fitting problem in test-based program synthesis and automatic program repair, which is a consequence of the incomplete specification, the limited tests or training examples, that the program synthesis or repair algorithm has to learn from. Recently, test generation techniques have been successfully employed to augment existing specifications of intended program behavior, to improve the generalizability of program synthesis and repair. Inspired by these approaches, in this paper, we propose a technique that re-purposes software testing methods, specifically mutation-based fuzzing, to augment the training data of DNNs, with the objective of enhancing their robustness. Our technique casts the DNN data augmentation problem as an optimization problem. It uses genetic search to generate the most suitable variant of an input data to use for training the DNN, while simultaneously identifying opportunities to accelerate training by skipping augmentation in many instances. We instantiate this technique in two tools, SENSEI and SENSEI-SA, and evaluate them on 15 DNN models spanning 5 popular image data-sets. Our evaluation shows that SENSEI can improve the robust accuracy of the DNN, compared to the state of the art, on each of the 15 models, by upto 11.9% and 5.5% on average. Further, SENSEI-SA can reduce the average DNN training time by 25%, while still improving robust accuracy.

Thu 9 Jul

Displayed time zone: (UTC) Coordinated Universal Time change

00:00 - 01:00
P14-TestingTechnical Papers / Software Engineering in Practice at Goguryeo
Chair(s): Shin Yoo Korea Advanced Institute of Science and Technology
00:00
12m
Talk
Seenomaly: Vision-Based Linting of GUI Animation Effects Against Design-Don’t GuidelinesTechnical
Technical Papers
Dehai Zhao Australian National University, Zhenchang Xing Australia National University, Chunyang Chen Monash University, Xiwei Xu Data 61, Liming Zhu CSIRO's Data61 and UNSW, Guoqiang Li Shanghai Jiao Tong University, Jinshui Wang School of Information Science and Engineering, Fujian University of Technology, Fuzhou, China
00:12
12m
Talk
Fuzz Testing based Data Augmentation to Improve Robustness of Deep Neural NetworksTechnical
Technical Papers
Xiang Gao National University of Singapore, Singapore, Ripon Saha Fujitsu Laboratories of America, Inc., Mukul R. Prasad Fujitsu Laboratories of America, Inc, Abhik Roychoudhury National University of Singapore, Singapore
00:24
12m
Talk
Modeling and Ranking Flaky Tests at AppleSEIP
Software Engineering in Practice
Emily Kowalczyk Apple Inc., Karan Nair Apple, Zebao Gao Apple, Leopold Silberstein Apple Inc., Teng Long Apple, Atif Memon Apple Inc.
00:36
12m
Talk
Testing File System Implementations on Layered ModelsTechnicalArtifact Available
Technical Papers
Dongjie Chen Nanjing University, Yanyan Jiang Nanjing University, Chang Xu Nanjing University, Xiaoxing Ma Nanjing University, Jian Lu Nanjing University
00:48
12m
Talk
A Cost-efficient Approach to Building in Continuous IntegrationTechnical
Technical Papers
Xianhao Jin Virginia Tech, USA, Francisco Servant Virginia Tech
Pre-print