"Machine Learning" is the theory of fitting models to data while simultaneously accounting for the computational complexity of the learning algorithm and the model's performance on future data. Its primary goal is to produce learning procedures that are computationally efficient and perform well (near optimal) on future data. The Machine Learning team in CCS-3 is working to advance the foundations of Machine Learning (ML) by developing rigorous explanations for the success of modern predictor design methods such as Support Vector Machines (SVMs) and Boosting, and expanding the theoretical framework used to analyze learning methods (e.g. tools for estimating and bounding generalization error and computational complexity). We also facilitate the use of modern machine learning methods in projects of national interest. Examples of programs at LANL that contain elements of Machine Learning include stockpile stewardship, functional genomics, nonproliferation, radiography, fraud detection, and computer security.

Publications