Graph resistance and learning from pairwise comparisons
dc.contributor.author | Hendrickx, Julien M. | en_US |
dc.contributor.author | Olshevsky, Alex | en_US |
dc.contributor.author | Saligrama, Venkatesh | en_US |
dc.date.accessioned | 2019-08-30T12:43:26Z | |
dc.date.available | 2019-08-30T12:43:26Z | |
dc.date.issued | 2019 | |
dc.identifier | http://arxiv.org/abs/1902.00141v2 | |
dc.identifier.citation | Julien M Hendrickx, Alex Olshevsky, Venkatesh Saligrama. "Graph Resistance and Learning from Pairwise Comparisons." Julien Hendrickx, Alexander Olshevsky, Venkatesh Saligrama ; Proceedings of the 36th International Conference on Machine Learning, PMLR 97:2702-2711, 2019. | |
dc.identifier.uri | https://hdl.handle.net/2144/37589 | |
dc.description.abstract | We consider the problem of learning the qualities of a collection of items by performing noisy comparisons among them. Following the standard paradigm, we assume there is a fixed “comparison graph” and every neighboring pair of items in this graph is compared k times according to the Bradley-Terry-Luce model (where the probability than an item wins a comparison is proportional the item quality). We are interested in how the relative error in quality estimation scales with the comparison graph in the regime where k is large. We show that, asymptotically, the relevant graph-theoretic quantity is the square root of the resistance of the comparison graph. Specifically, we provide an algorithm with relative error decay that scales with the square root of the graph resistance, and provide a lower bound showing that (up to log factors) a better scaling is impossible. The performance guarantee of our algorithm, both in terms of the graph and the skewness of the item quality distribution, significantly outperforms earlier results. | en_US |
dc.subject | Machine learning | en_US |
dc.subject | Mathematics | en_US |
dc.subject | Computer science | en_US |
dc.title | Graph resistance and learning from pairwise comparisons | en_US |
dc.type | Article | en_US |
dc.description.version | Published version | en_US |
pubs.elements-source | arxiv | en_US |
pubs.notes | 15 pages, including 5 pages supplementary material | en_US |
pubs.notes | Embargo: Not known | en_US |
pubs.organisational-group | Boston University | en_US |
pubs.organisational-group | Boston University, College of Engineering | en_US |
pubs.organisational-group | Boston University, College of Engineering, Department of Electrical & Computer Engineering | en_US |
dc.identifier.orcid | 0000-0002-0675-2268 (Saligrama, Venkatesh) | |
dc.identifier.mycv | 423994 |
This item appears in the following Collection(s)