Wasserstein Barycenter Model Ensembling

Abstract

In this paper we propose to perform model ensembling in a multiclass or a multilabel learning setting using Wasserstein (W.) barycenters. Optimal transport metrics, such as the Wasserstein distance, allow incorporating semantic side information such as word embeddings. Using W. barycenters to find the consensus between models allows us to balance confidence and semantics in finding the agreement between the models. We show applications of Wasserstein ensembling in attribute-based classification, multilabel learning and image captioning generation. These results show that the W. ensembling is a viable alternative to the basic geometric or arithmetic mean ensembling.

arXiv commentICLR 2019

Cited by in corpus (9)

• Ground Metric Learning on Graphs
• Evaluating the Disentanglement of Deep Generative Models through Manifold Topology
• Limit Distribution Theory for the Smooth 1-Wasserstein Distance with Applications
• Estimating Barycenters of Measures in High Dimensions
• Statistical Optimal Transport posed as Learning Kernel Embedding
• Fixed Support Tree-Sliced Wasserstein Barycenter
• Non-Asymptotic Performance Guarantees for Neural Estimation of $\mathsf{f}$-Divergences
• Riemannian-geometric generalizations of quantum fidelities and Bures-Wasserstein distance
• Variational Wasserstein Barycenters with c-Cyclical Monotonicity