Adversarial Controls for Scientific Machine Learning

• DOI: 10.1021/acschembio.8b00881
• Publication Date: 2018-10-19T04:03:02Z
• Authors: Kangway V. Chuang; Michael J. Keiser
• Keywords: Leverage (statistics); Cornerstone

New machine learning methods to analyze raw chemical and biological data are now widely accessible as open-source toolkits. This positions researchers leverage powerful, predictive models in their own domains. We caution, however, that the application of experimental research merits careful consideration. Machine algorithms readily exploit confounding variables artifacts instead relevant patterns, leading overoptimistic performance poor model generalization. In parallel strong control experiments remain a cornerstone research, we advance concept adversarial controls for scientific learning: design exacting purposeful ensure arises from meaningful models.