A Review and Refinement of Surprise Adequacy

TL;DR

This paper reviews and refines the computation of Surprise Adequacy (SA) for Deep Learning testing, introducing optimized algorithms that significantly reduce evaluation time and analyzing SA's effectiveness and sensitivity in out-of-distribution detection.

Contribution

It presents a performance-optimized implementation of SA, refined variants for faster evaluation, and an empirical study on MNIST highlighting SA's capabilities and sensitivity issues.

Findings

Refined SA variants are substantially faster with comparable results.

Optimized implementation reduces evaluation time by up to 97%.

SA can be highly sensitive to non-determinism in DNN training.

Abstract

Surprise Adequacy (SA) is one of the emerging and most promising adequacy criteria for Deep Learning (DL) testing. As an adequacy criterion, it has been used to assess the strength of DL test suites. In addition, it has also been used to find inputs to a Deep Neural Network (DNN) which were not sufficiently represented in the training data, or to select samples for DNN retraining. However, computation of the SA metric for a test suite can be prohibitively expensive, as it involves a quadratic number of distance calculations. Hence, we developed and released a performance-optimized, but functionally equivalent, implementation of SA, reducing the evaluation time by up to 97\%. We also propose refined variants of the SA omputation algorithm, aiming to further increase the evaluation speed. We then performed an empirical study on MNIST, focused on the out-of-distribution detection capabilities of SA, which allowed us to reproduce parts of the results presented when SA was first released. The experiments show that our refined variants are substantially faster than plain SA, while producing comparable outcomes. Our experimental results exposed also an overlooked issue of SA: it can be highly sensitive to the non-determinism associated with the DNN training procedure.