CS364B: Frontiers in Mechanism Design Lecture #9: MIDR Mechanisms via Scaling Algorithms∗
نویسنده
چکیده
Theorem 1.1 If k ≥ c logm 2 for a sufficiently large constant c, then there is a randomized algorithm with expected polynomial running time that, with probability 1, outputs a feasible allocation with welfare at least 1− times the maximum possible. ∗ c ©2014, Tim Roughgarden. †Department of Computer Science, Stanford University, 462 Gates Building, 353 Serra Mall, Stanford, CA 94305. Email: [email protected].
منابع مشابه
Cs364b: Frontiers in Mechanism Design Lecture #7: Submodular Valuations
With this lecture we commence Part II of the course. We continue to pursue very strong incentive guarantees (i.e., DSIC mechanisms) in scenarios where the underlying welfaremaximization problem is NP-hard. This turns out to be a very difficult quest; the goal of this part of the course is to highlight the challenges involved and survey what few design techniques and positive results are known. ...
متن کاملCS364B: Frontiers in Mechanism Design Lecture #1: Ascending and Ex Post Incentive Compatible Mechanisms∗
These twenty lectures cover advanced topics in mechanism design. They assume familiarity with some of the material covered in the instructor’s CS364A course — specifically, lectures 2–4 and 7–9. Recall that mechanism design is the “science of rule-making.” The goal is to understand how to design systems with strategic participants — autonomous decision-makers whose objectives are generally diff...
متن کاملCS364B: Frontiers in Mechanism Design Lecture #10: Coverage Valuations and Convex Rounding∗
Recall that the submodularity condition in (1) is a form of diminishing returns, but is strictly weaker than the gross substitute condition. If we ignore incentives and assume straightforward bidding, the Kelso-Crawford auction achieves a 1 2 -approximation of the optimal welfare. Another 1 2 -approximation is given in Exercise #25. Are there DSIC mechanisms with equally good guarantees? While ...
متن کاملCS364B: Frontiers in Mechanism Design Lecture #20: Characterization of Revenue-Maximizing Auctions∗
Recall what we are shooting for: a multi-parameter analog of Myerson’s theory of revenuemaximizing optimal auctions. Single-parameter optimal mechanisms are virtual welfare maximizers. We’ve seen that multi-parameter optimal mechanisms are generally randomized. It would seem that the coolest statement that could be true is: revenue-maximizing optimal auctions for multi-parameter problems are al...
متن کاملCS364B: Frontiers in Mechanism Design Lecture #14: The Price of Anarchy in Simple Auctions
With this lecture we commence the fourth part of the course. In the last two parts we took incentive constraints, like DSIC or BIC, as a hard constraint, and subject to this designed the best mechanisms possible. Most of the mechanisms discussed were quite complex — interesting as proofs of concept of what is possible in principle, but not suitable for actual use. This part of the course insist...
متن کامل