Sherali--Adams Proof System || @ CMU || Lecture 21b of CS Theory Toolkit

Sherali--Adams Proof System || @ CMU || Lecture 21b of CS Theory Toolkit

🎙 Ryan O'Donnell 👥 14K 📅 June 19, 2020 ⏱ 36 min 👁 1K 📄 lecture 🧭 2026-08-17
Available in: English (current) Français

Keywords

Sherali-Adamsproof systemLP relaxationmax-2SATmax-cutsemialgebraic proofsautomatizability

Summary

This lecture introduces the Sherali-Adams proof system, a hierarchy of linear programming (LP) relaxations for combinatorial optimization problems. The instructor begins with a motivating example of max-2SAT, showing how a degree-2 polynomial inequality can capture the objective and how adding axioms for non-negativity of certain terms allows deriving an upper bound. He then generalizes to a proof system with degree-k polynomial inequalities, where axioms are all true inequalities involving at most k variables. The system is sound, but incomplete for k < n. It is static, meaning derivations can be done in one step, and it is automatizable: if a derivation exists, it can be found efficiently in time polynomial in the number of axioms (n^O(k)). The lecture also discusses the hierarchy’s increasing strength with k and contrasts it with dynamic proof systems like cutting planes. The presentation is clear and rigorous, with examples and references to further resources.

149 words

Critical Evaluation

Value of the Information & Strength of the Argument

The lecture provides a solid introduction to the Sherali-Adams proof system, explaining its motivation, formal definition, and key properties. The argumentation is well-structured: it starts with concrete examples (max-2SAT and max-cut) to illustrate the need for a more powerful proof system, then formalizes the system and discusses its soundness, completeness, and automatizability. The instructor also explains why the system is automatizable by showing that any derivable inequality can be expressed as a non-negative linear combination of a finite set of axioms. The presentation is rigorous and accessible to a graduate-level audience, with clear explanations of technical concepts.

Scientific Rigor, Source Quality, Title Accuracy

The lecture is scientifically rigorous, with a clear and accurate presentation of the Sherali-Adams proof system. The instructor references a relevant survey by Fleming, Kothari, and Pitassi, which is a reputable source in the field. The title accurately describes the content, and the lecture is part of a well-structured graduate course. The presentation is self-contained, with no apparent errors or misleading statements. The description includes links to the instructor’s homepage and course materials, which are appropriate resources for further study.

192 words

Title / Content Match

The title accurately reflects the content: a lecture on the Sherali-Adams proof system, part of a CS theory course.

Quality & Reliability

8/10

Lecture by a recognized expert in theoretical computer science, part of a graduate course at Carnegie Mellon. The content is rigorous, well-structured, and includes references to a relevant survey. The presentation is clear and technically accurate, with appropriate caveats about completeness and automatizability.

Key Moments

Cited Sources

  • Ryan O'Donnell's homepage — Instructor's academic page, providing background and related materials.
  • Course homepage on Diderot — Course page for CS Theory Toolkit, where lecture notes and resources are available.
  • Rebecca Kiger Photography — Photographer credited for the thumbnail image.

Concurring Sources

Contribution & Novelties

This lecture provides a clear and accessible introduction to the Sherali-Adams proof system, a fundamental tool in theoretical computer science for designing LP relaxations. It explains the motivation, formal definition, and key properties, including soundness, incompleteness, and automatizability. The lecture also highlights the hierarchy’s increasing strength with k and its applications to combinatorial optimization problems like max-2SAT and max-cut.

Pour aller plus loin :

  • Sherali-Adams hierarchy — Overview of the hierarchy and its applications.
  • Semialgebraic Proofs and Efficient Algorithm Design — Survey by Fleming, Kothari, and Pitassi, referenced in the lecture.
  • Lovász–Schrijver hierarchy — Related proof system hierarchy, useful for comparison.

101 words

Radar Profile

The radar profile shows high scores in information quantity, quality, and technical level, with slightly lower but still strong reliability. This indicates a dense, technically rigorous lecture that is well-suited for an advanced audience, with a solid foundation in the subject matter.

Reliability 8/10