CCLVQ trains n expert networks with a winner-takes-all loss plus a classifier, so the ensemble approximates the conditional law of Y given X in Wasserstein distance.
Relaxing bijectivity constraints with continuously indexed normalising flows
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Conditional Distribution Quantization in Machine Learning
CCLVQ trains n expert networks with a winner-takes-all loss plus a classifier, so the ensemble approximates the conditional law of Y given X in Wasserstein distance.