Book
Open access
Jul 2026
Evolutionary Feature Construction for Tabular Transformers via Gradient-Enhanced Attention Rollout
This work introduces GEAR, a gradient-enhanced attention rollout that preserves the full interaction matrix between input tokens and uses its input-input block as an interaction prior for a trained tabular Transformer to guide a validity-first genetic programming search.
H. Byun, Jieun Yook, Hanbin Koo et al.
· GECCO Companion · 0 citations