L 0 Feature Selection via Monte Carlo Tree Search in High-Dimensional Financial Regression

Name: L 0 Feature Selection via Monte Carlo Tree Search in High-Dimensional Financial Regression
Uploaded: 2026-10-19T01:58:28.933606+00:00
Duration: 430 s
Description: The method uses Monte Carlo Tree Search with a policy adaptation to find sparse feature subsets for financial regression. It outperforms Lasso by finding better combinations of variables, achieving an ensemble out-of-sample R² of 0.6961 compared to Lasso's 0.6313, proving the search method identifies superior predictive features.

The method uses Monte Carlo Tree Search with a policy adaptation to find sparse feature subsets for financial regression. It outperforms Lasso by finding better combinations of variables, achieving an ensemble out-of-sample R² of 0.6961 compared to Lasso's 0.6313, proving the search method identifies superior predictive features.

Open the full episode