Validating a Deep Learning Framework by Metamorphic Testing
Authors:
- Junhua DingJDJunhua Ding
- Xiaojun KangXKXiaojun Kang
- Xin-Hua HuXHXin-Hua Hu
- 1 May 2017
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 28-34
Abstract
Deep learning has become an important tool for image classification and natural language processing. However, the effectiveness of deep learning is highly dependent on the quality of the training data as well as the net model for the learning. The training data set for deep learning normally is fairly large, and the net model is pretty complex. It is necessary to validate the deep learning framework including the net model, executing environment, and training data set before it is used for any applications. In this paper, we propose an approach for validating the classification accuracy of a deep learning framework that includes a convolutional neural network, a deep learning executing environment, and a massive image data set. The framework is first validated with a classifier built on support vector machine, and then it is tested using a metamorphic validation approach. The effectiveness of the approach is demonstrated by validating a deep learning classifier for automated classification of biology cell images. The proposed approach can be used for validating other deep learning framework for different applications.Keywords
This publication has 15 references indexed in Scilit:
- A Survey on Metamorphic TestingIEEE Transactions on Software Engineering, 2016
- Google AI algorithm masters ancient game of GoNature, 2016
- Feature Selection and Analysis of Diffraction Images2015
- Polarization imaging and classification of Jurkat T and Ramos B cells using a flow cytometerCytometry Part A, 2014
- Analysis of cellular objects through diffraction images acquired by flow cytometryOptics Express, 2013
- Representation Learning: A Review and New PerspectivesIEEE Transactions on Pattern Analysis and Machine Intelligence, 2013
- Development of a diffraction imaging flow cytometerOptics Letters, 2009
- A Tutorial on Support Vector Machines for Pattern RecognitionData Mining and Knowledge Discovery, 1998
- Pseudo-oracles for non-testable programs1981
- Textural Features for Image ClassificationIEEE Transactions on Systems, Man, and Cybernetics, 1973