en
0.25
0.5
0.75
1.25
1.5
1.75
2
Exact Bayesian Pairwise Preference Learning and Inference on the Uniform Convex Polytope
Published on 2012-01-243868 Views
In Bayesian approaches to utility learning from preferences, the objective is to infer a posterior belief distribution over an agent’s utility function based on previously observed agent preferences.
Related categories
Presentation
Exact Bayesian Pairwise Preference Learning and Inference on the Uniform Convex Polytope00:00
Pic00:06
Part I: Bayesian Pairwise Preference Learning (BPPL)00:25
Learning from Preferences00:43
Bayesian Decision Theory01:37
Preference Feedback Types02:44
Preference Learning Feedback04:01
Part II: Bayesian Pairwise Preference Learning (BPPL)04:53
Bayesian Utility Model05:13
Response Likelihood Model05:40
Geometric View06:26
Uniform Prior07:32
Inference Tasks08:20
Polynomial Case Representation09:11
Polynomial Case Operations: - 110:06
Polynomial Case Operations: - 210:25
Polynomial Case Operations: max - 111:13
Polynomial Case Operations: max - 211:28
BPPL Inference12:11
Marginalization: ∫ - 112:45
Marginalization: ∫ - 213:34
Definite Integral Evaluation15:10
That’s It!16:00
Case ® XADD16:35
Maintaining Compact Cases17:08
Compactness of (X)ADDs17:45
Binary Operations on (X)ADDs18:02
XADD Maximization18:35
Maintaining XADD Orderings19:11
Correcting XADD Ordering19:49
XADD Pruning20:21
Empirical Results21:07
Future Work – Exact BPPL21:59
Thank you!22:58