This repository contains code to reproduce experiments in Appendix C "Illustrative experiments on data regularization" of our paper "Reciprocal Learning" (NeurIPS 2024)
- R contains implementation of BPLS with PPP and alternative PLS methods to benchmark against
- benchmarking provides files for experiments, in order to reproduce results, see setup below
- data contains real-world data (banknote and breast cancer data) used in experiments
- experimental results and visualization thereof will be saved in plots and results
- R 4.2.0
- R 4.1.6
- R 4.0.3
on
- Linux Ubuntu 20.04
- Linux Debian 10
- Windows 11 Pro Build 22H2
First and foremost, clone this repo and please install all dependencies by sourcing this file.
The implementations of self-training with different selection criteria as detailed in the paper are in folder "R":
- Supervised Baseline
- Probability Score
- Predictive Variance
- PPP (Bayes-optimal)
- Regularized PPP (Bayes-optimal)
- Likelihood (max-max)
- Utilities for PPP
Experimental setups are in these files:
- experiments with likelihood (max-max)
- experiments with PPP (bayes-opt)
- experiments with regularized PPP (bayes-opt)
- experiments with supervised baseline
- experiments with predictive variance
- experiments with probability score
In order to reproduce results (and visualizations thereof) in Appendix C "Illustrative experiments on data regularization" of our paper "Reciprocal Learning", run
Important: Make sure you have folders results and plots where experimental results will be stored automatically. In addition, you can access them as object after completion of the experiments.
Additional illustrating experimental setups can now easily be created by modifying setup in benchmarks/experiments_banknote.R