Beyond the Worst-Case Analysis of Algorithms

Algorithms, but with more realistic lenses

Beyond the Worst-Case Analysis of Algorithms gathers a large body of work on algorithm analysis that looks past the familiar worst-case framework. Edited by Tim Roughgarden, the volume introduces several alternative ways to reason about performance, from parameterized analysis and instance optimality to smoothed analysis and semirandom models.

The result is a book that feels less like a single-threaded textbook and more like a map of a field: one that connects classic algorithmic questions with modern work in clustering, linear programming, sparse recovery, tensor methods, and machine learning. The chapters are designed as introductions, so the emphasis is on the main ideas, key results, and the models that make them work.

What the book covers

• Refinements of worst-case analysis, including parameterized algorithms and resource augmentation
• Deterministic models of data such as perturbation resilience, approximation stability, and sparse recovery
• Semirandom and random-order models, including stochastic block models and self-improving algorithms
• Smoothed analysis of local search, the simplex method, and multiobjective optimization
• Applications in machine learning and statistics, including robust high-dimensional statistics, tensor decompositions, topic models, and generalization in overparameterized models

A guided introduction to a broad field

This is a strong fit for readers who want an organized, chapter-by-chapter overview of modern algorithmic analysis without having to piece the subject together from scattered papers. The contributors include leading researchers, and the structure makes the material approachable for advanced students as well as working researchers.

If your catalog needs a serious, research-oriented algorithms title with strong relevance to optimization and machine learning, this book is an excellent addition.