Discovering Boundary Values of Feature-based Machine Learning Classifiers through Exploratory Datamorphic Testing

TL;DR

This paper introduces exploratory datamorphic testing strategies and algorithms for identifying class boundaries in machine learning classifiers, enhancing understanding of model behavior through formal proofs and experimental evaluation.

Contribution

It presents novel testing strategies and algorithms within the datamorphism testing framework for discovering class boundaries in ML models, with formal correctness proofs and empirical evaluation.

Findings

Algorithms effectively discover class boundaries

Formal proofs ensure algorithm correctness

Experimental results demonstrate practical utility

Abstract

Testing has been widely recognised as difficult for AI applications. This paper proposes a set of testing strategies for testing machine learning applications in the framework of the datamorphism testing methodology. In these strategies, testing aims at exploring the data space of a classification or clustering application to discover the boundaries between classes that the machine learning application defines. This enables the tester to understand precisely the behaviour and function of the software under test. In the paper, three variants of exploratory strategies are presented with the algorithms implemented in the automated datamorphic testing tool Morphy. The correctness of these algorithms are formally proved. Their capability and cost of discovering borders between classes are evaluated via a set of controlled experiments with manually designed subjects and a set of case studies with real machine learning models.