FoGS filters a pool of samples from four tabular generators using an ensemble of survival models to produce synthetic data that improves C-index by +2.17 and IBS by +0.67 on average across 16 datasets compared with unfiltered baselines.
Adversarial random forests for density estimation and generative modeling
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Filtered Mixture-of-Generators for Fully Synthetic Survival Training
FoGS filters a pool of samples from four tabular generators using an ensemble of survival models to produce synthetic data that improves C-index by +2.17 and IBS by +0.67 on average across 16 datasets compared with unfiltered baselines.