New KINDER method targets fair representation learning via prototype-space erasure
Researchers introduced KINDER, a kernel-based approach for concept erasure that removes sensitive attributes from learned representations to improve machine learning fairness.
Read the full story at arXiv cs.LG ↗
Timeline · 1 report
- 2026-11-03 05:00 · arXiv cs.LG
KINDER: Kernel-based Independence for Fair Representation Learning via Prototype-space Erasure