Feature Selection for Latent Factor Models

Name: Feature Selection for Latent Factor Models
Uploaded: 2026-10-22T10:22:19.863629+00:00
Duration: 360 s
Description: The paper introduces an SNR-based feature selection method for latent factor models like LFA and PPCA. It selects features for each class independently using a signal-to-noise ratio criterion, offering theoretical guarantees of true feature recovery. This approach outperforms existing methods on standard datasets and is useful in class-incremental learning.

The paper introduces an SNR-based feature selection method for latent factor models like LFA and PPCA. It selects features for each class independently using a signal-to-noise ratio criterion, offering theoretical guarantees of true feature recovery. This approach outperforms existing methods on standard datasets and is useful in class-incremental learning.

Open the full episode