Activity Number:
|
405
- Student Paper Award and Chambers Statistical Software Award
|
Type:
|
Topic Contributed
|
Date/Time:
|
Wednesday, August 5, 2020 : 1:00 PM to 2:50 PM
|
Sponsor:
|
Section on Statistical Computing
|
Abstract #309920
|
|
Title:
|
End-to-End Statistical Learning, with or without Labels
|
Author(s):
|
Corinne Jones* and Vincent Roulet and Zaid Harchaoui
|
Companies:
|
University of Washington and University of Washington and University of Washington
|
Keywords:
|
discriminative clustering;
unsupervised learning;
semi-supervised learning;
representation learning
|
Abstract:
|
We introduce an approach that allows one to learn a feature representation and perform clustering of unlabeled data. The approach can also leverage any amount of additional labeled data in order to boost the statistical performance. The proposed method is based on a semi-implicit stochastic optimization algorithm and an entropy-regularized optimal transport algorithm. A numerical illustration on a real dataset shows the promise of the proposed approach.
|
Authors who are presenting talks have a * after their name.