Accepted Paper: Variational Inference from Ranked Samples with Features
Back to list of accepted papers
- Session 6: Supervised Learning -- Day 3 (Nov.19), poster session: 11:30-14:00, talks: 14:10-15:25 (5th floor Hall 2)
- Poster number: Tue33
- Download paper
Authors
Yuan Guo (Northeastern University); Jennifer Dy (Northeastern); Deniz Erdogmus (Northeastern University); Jayashree Kalpathy-Cramer (MGH/Harvard Medical School); Susan Ostmo (Oregon Health & Science University); J. Peter Campbell (Oregon Health & Science University); Michael F. Chiang (Oregon Health & Science University); Stratis Ioannidis (Northeastern University)
Abstract
In many supervised learning settings,elicited labels comprise pairwise comparisons or rankings of samples. We propose a Bayesian inference model for ranking datasets, allowing us to take a probabilistic approach toranking inference. Our probabilistic assumptions are motivated by, and consistent with, the so-called Plackett-Luce model. We propose a variational inference method to extract a closed-form Gaussian posterior distribution, and show experimentally that the closed form posterior distribution yields a more reliable ranking prediction over prediction via point estimates.