This project intends to focus on several aspects of this development: self-improving algorithms; online data reconstruction; sublinear algorithms; dimension reduction; low en-tropy data structures, nonuniformly priced computation; and data analysis. The scope of the proposed research is intentionally wide-ranging. Indeed, it is the PI's belief, backed by his preliminary investigations, that progress on these topics will rely on the emergence of common threads and broadly applicable methods. The intellectual merit of this proposal is to lay down the foundations for a systematic study of data-powered algorithms. The eort will involve theoretical investigations coupled with extensive computer experimentation. The principal beneciaries of data-powered algorithms come from engineering and the natural sciences, and this is where the broader impact of this proposal will be sought; specically, via the PI's ongoing research collaborations with biologists and physicists and, on the educational front, his joint effort with them to develop a year-long algorithms-based course for quantitative scientists.
