Hypernetworks for Dynamic Feature Selection

Name: Hypernetworks for Dynamic Feature Selection
Uploaded: 2026-10-03T14:30:09.126510+00:00
Duration: 546 s
Description: Hyper-DFS is a dynamic feature selection method that uses a hypernetwork to generate specific classifier parameters for every feature subset encountered during training. It overcomes limitations of existing methods by using Set Transformers to encode feature subsets, allowing it to generalize better on both synthetic and real data.

Hyper-DFS is a dynamic feature selection method that uses a hypernetwork to generate specific classifier parameters for every feature subset encountered during training. It overcomes limitations of existing methods by using Set Transformers to encode feature subsets, allowing it to generalize better on both synthetic and real data.

Open the full episode