Winner-takes-all learners are geometry-aware conditional density estimators

Victor Letzelter    David Perera    Cédric Rommel    Mathieu Fontaine    Slim Essid    Gaël Richard    Patrick Pérez

ICML 2024

Abstract

Winner-takes-all training is a simple learning paradigm, which handles ambiguous tasks by predicting a set of plausible hypotheses. Recently, a connection was established between Winner-takes-all training and centroidal Voronoi tessellations, showing that, once trained, hypotheses should quantize optimally the shape of the conditional distribution to predict. However, the best use of these hypotheses for uncertainty quantification is still an open question. In this work, we show how to leverage the appealing geometric properties of the Winner-takes-all learners for conditional density estimation, without modifying its original training scheme. We theoretically establish the advantages of our novel estimator both in terms of quantization and density estimation, and we demonstrate its competitiveness on synthetic and real-world datasets, including audio data.

BibTeX

@inproceedings{letzelter2024winner,
  title={Winner-takes-all learners are geometry-aware conditional density estimators},
  author={Letzelter, Victor and Perera, David and Rommel, C{\'e}dric and Fontaine, Mathieu and Essid, Slim and Richard, Gael and P{\'e}rez, Patrick},
  booktitle={International Conference on Machine Learning},
  year={2024}
}