Novel algorithms for efficient learning of distributional regression trees optimized for CRPS and WIS losses via heaps, balanced trees, and Fenwick trees, with competitive performance and conformal prediction applications.
Fenchel- Y oung losses with skewed entropies for class-posterior probability estimation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Efficient distributional regression trees learning algorithms for calibrated non-parametric probabilistic forecasts
Novel algorithms for efficient learning of distributional regression trees optimized for CRPS and WIS losses via heaps, balanced trees, and Fenwick trees, with competitive performance and conformal prediction applications.