Summary
Many important artificial intelligence computing systems rely on algorithms to make decisions about scheduling, routing, and the use of limited resources. Traditional algorithms are dependable because their performance can be guaranteed in all cases, but they often cannot take advantage of recurring patterns in data that could make them faster or more effective. By contrast, machine learning methods can detect useful patterns, but their guidance may become unreliable when conditions change or when the learned estimates are inaccurate. This project develops new algorithmic methods that combine
