With the ever‐growing adoption of artificial intelligence (AI)‐based systems, the carbon footprint of AI is no longer negligible. AI researchers and practitioners are therefore urged to hold themselves accountable for the carbon emissions of the AI models they design and use. This led in recent years to the appearance of researches tackling AI environmental sustainability, a field referred to as Green AI. Despite the rapid growth of interest in the topic, a comprehensive overview of Green AI research is to date still missing. To address this gap, in this article, we present a systematic review of the Green AI literature. From the analysis of 98 primary studies, different patterns emerge. The topic experienced a considerable growth from 2020 onward. Most studies consider monitoring AI model footprint, tuning hyperparameters to improve model sustainability, or benchmarking models. A mix of position papers, observational studies, and solution papers are present. Most papers focus on the training phase, are algorithm‐agnostic or study neural networks, and use image data. Laboratory experiments are the most common research strategy. Reported Green AI energy savings go up to 115%, with savings over 50% being rather common. Industrial parties are involved in Green AI studies, albeit most target academic readers. Green AI tool provisioning is scarce. As a conclusion, the Green AI research field results to have reached a considerable level of maturity. Therefore, from this review emerges that the time is suitable to adopt other Green AI research strategies, and port the numerous promising academic results to industrial practice.This article is categorized under: Technologies > Machine Learning

Test suite reduction approaches aim at decreasing software regression testing costs by selecting a representative subset from large-size test suites. Most existing techniques are too expensive for handling modern massive systems and moreover depend on artifacts, such as code coverage metrics or specification models, that are not commonly available at large scale. We present a family of novel very efficient approaches for similaritybased test suite reduction that apply algorithms borrowed from the big data domain together with smart heuristics for finding an evenly spread subset of test cases. The approaches are very general since they only use as input the test cases themselves (test source code or command line input).We evaluate four approaches in a version that selects a fixed budget B of test cases, and also in an adequate version that does the reduction guaranteeing some fixed coverage. The results show that the approaches yield a fault detection loss comparable to state-of-the-art techniques, while providing huge gains in terms of efficiency. When applied to a suite of more than 500K real world test cases, the most efficient of the four approaches could select B test cases (for varying B values) in less than 10 seconds.

Context: Flaky tests plague regression testing in Continuous Integration environments by slowing down change releases and wasting testing time and effort. Despite the growing interest in mitigating the burden of test flakiness, how to efficiently and effectively detect flaky tests is still an open problem. Objective: In this study, we present and evaluate FLAST, an approach designed to statically predict test flakiness. FLAST leverages vector-space modeling, similarity search, dimensionality reduction, and k-Nearest Neighbor classification in order to timely and efficiently detect test flakiness. Method: In order to gain insights into the efficiency and effectiveness of FLAST, we conduct an empirical evaluation of the approach by considering 13 real-world projects, for a total of 1,383 flaky and 26,702 non-flaky tests. We carry out a quantitative comparison of FLAST with the state-of-the-art methods to detect test flakiness, by considering a balanced dataset comprising 1,402 real-world flaky and as many non-flaky tests. Results: From the results we observe that the effectiveness of FLAST is comparable with the state-of-the-art, while providing considerable gains in terms of efficiency. In addition, the results demonstrate how by tuning the threshold of the approach FLAST can be made more conservative, so to reduce false positives, at the cost of missing more potentially flaky tests. Conclusion: The collected results demonstrate that FLAST provides a fast, low-cost and reliable approach that can be used to guide test rerunning, or to gate the inclusion of new potentially flaky tests.

Context: Tangled commits are changes to software that address multiple concerns at once. For researchers interested in bugs, tangled commits mean that they actually study not only bugs, but also other concerns irrelevant for the study of bugs. Objective: We want to improve our understanding of the prevalence of tangling and the types of changes that are tangled within bug fixing commits. Methods: We use a crowd sourcing approach for manual labeling to validate which changes contribute to bug fixes for each line in bug fixing commits. Each line is labeled by four participants. If at least three participants agree on the same label, we have consensus. Results: We estimate that between 17% and 32% of all changes in bug fixing commits modify the source code to fix the underlying problem. However, when we only consider changes to the production code files this ratio increases to 66% to 87%. We find that about 11% of lines are hard to label leading to active disagreements between participants. Due to confirmed tangling and the uncertainty in our data, we estimate that 3% to 47% of data is noisy without manual untangling, depending on the use case. Conclusion: Tangled commits have a high prevalence in bug fixes and can lead to a large amount of noise in the data. Prior research indicates that this noise may alter results. As researchers, we should be skeptics and assume that unvalidated data is likely very noisy, until proven otherwise.