afa-ssl: Code for label-efficient active feature acquisition with a self-supervised pretrained predictor

Code accompanying the manuscript "Self-supervised pretraining of the predictor for label-efficient active feature acquisition: application to cancer gene expression" by Mahdi Sabet Kish and Mehran Yazdi, submitted to Knowledge-Based Systems.

This release contains the full pipeline used for the experiments in the paper: data loaders, models, acquisition criteria, baselines (EDDI, DIME), the job runner, and the scripts that produce the tables and figures. See README.md for installation and data/README.md for downloading the datasets.