Modern artificial intelligence tools rely on the use of machine learning technology, and this technology would not be possible without the availability of efficient algorithms for solving mathematical optimization problems. Lehigh ISE is the home of a unique concentration of experts in mathematical optimization with multiple faculty members working to push the boundaries of algorithms for solving optimization problems arising in machine learning and various other data science applications.
Lehigh ISE faculty member Frank E. Curtis, in collaboration with Katya Scheinberg of Cornell University, have been awarded a three-year half-million-dollar award from the National Science Foundation (CISE Directorate, CCF Division, Algorithmic Foundations program) to design a unified framework for analyzing probabilistic algorithms for solving stochastic optimization problems. One of the disadvantages of contemporary algorithms is that they need to undergo extensive tuning procedures before they can be employed effectively in practice. The proposed framework of Curtis and Scheinberg, on the other hand, shows how adaptive “self-tuning” algorithms can be designed so that they possess strong theoretical guarantees and significantly reduce the overall computational time needed to solve challenging real-world learning problems.
Frank: "NSF’s Algorithmic Foundations program has provided strong, long-lasting support for our research groups’ work on the design and analysis of next-generation algorithms for solving stochastic optimization problems arising in critical areas such as machine learning. We’re extremely grateful for this support and are excited to share our work with the community of researchers and practitioners who aim to solve challenging data-driven learning problems with algorithms that are efficient in terms of time and computation.”