While Professor Sherwood's primary research focus is on the creation of efficient and effective computing systems, his history with ML goes back to 2004 with work appearing in the Journal of Machine Learning Research.  Since that time he has continued to explore learning based approaches to optimize the time-varying behavior of software, worked to establish privacy bounds on the release of information useful for training and optimization, and led the construction of extremely energy efficient methods for classification using both new logic encodings and superconducting materials.