Provable FDR Control for Deep Feature Selection: Deep MLPs and Beyond

Name: Provable FDR Control for Deep Feature Selection: Deep MLPs and Beyond
Uploaded: 2026-10-02T23:15:28.127796+00:00
Duration: 648 s
Description: This framework uses deep neural networks to select features while guaranteeing control over false discovery rate (FDR). It provides a theoretical foundation based on multi-index data models, proving that feature importance statistics can be used to select variables with a controlled error rate. This method works across various deep learning architectures and training methods.

This framework uses deep neural networks to select features while guaranteeing control over false discovery rate (FDR). It provides a theoretical foundation based on multi-index data models, proving that feature importance statistics can be used to select variables with a controlled error rate. This method works across various deep learning architectures and training methods.

Open the full episode