Understanding when and why algorithms work is a fundamental challenge. For problems ranging from clustering to linear programming to neural networks there are significant gaps between empirical performance and prediction based on traditional worst-case analysis. The book introduces exciting new methods for assessing algorithm performance.
