Information-Geometric Set Embeddings (IGSE): From Sets to Probability Distributions
arXiv:1911.12463
Abstract
This letter introduces an abstract learning problem called the "set embedding": The objective is to map sets into probability distributions so as to lose less information. We relate set union and intersection operations with corresponding interpolations of probability distributions. We also demonstrate a preliminary solution with experimental results on toy set embedding examples.
arXiv commentTo be presented at Sets & Partitions (NeurIPS 2019 workshop)
References in corpus (5)
- Distributed Representations of Words and Phrases and their Compositionality
- Auto-Encoding Variational Bayes
- Multiple Instance Learning: A Survey of Problem Characteristics and Applications
- On a generalization of the Jensen-Shannon divergence and the JS-symmetrization of distances relying on abstract means
- Fisher information distance: a geometrical reading