Reproduce “Learning Decision Trees and Forests with Algorithmic Recourse” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Kentaro Kanamori, Takuya Takagi, Ken Kobayashi, Yuichi Ike
Researchers developed Recourse-Aware Classification Tree (RACT), an algorithm that learns decision trees and forests inherently guaranteeing low-cost algorithmic recourse actions for a high proportion of instances. RACT delivers significantly higher recourse ratios and lower-cost, more plausible actions compared to baselines, while preserving predictive accuracy and computational efficiency.
Key insights
- Integrating recourse awareness directly into the tree-building process, rather than relying on post-hoc extraction, is crucial for inherently guaranteeing recourse actions.
- A tunable hyperparameter effectively balances the trade-off between predictive accuracy and the proportion of instances guaranteed recourse, offering flexible model design.
- Recourse-aware tree learning can maintain computational efficiency comparable to standard tree algorithms by leveraging techniques from adversarial training for split-finding.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.