Skip to content

A sparsified AutoEncoder to solve Semi-Supervised classification tasks

Notifications You must be signed in to change notification settings

CyprienGille/Semi-Supervised-AutoEncoder

Repository files navigation

Semisupervised Autoencoder

This repository contains the code from :

A new Semi-supervised classification method using a supervised autoencoder for biomedical applications, Gille C. and Guyard F. and Barlaud M., (2022). https://arxiv.org/abs/2208.10315 .

Repository contents

  • script_semisupervised.py : This is the main script used to produce the results shown in the paper. It generates plots in the plots directory, and saves results (metrics, losses...) as CSVs in the results_semi folder. All parameters are tunable near the start of the script.
  • param_plots : This is a helper script to reproduce the plots from Figures 2 and 3 of the aforementioned paper.
  • script_eta_optimization.py : This script is used to find the optimal sparsification parameter $\eta$ either by dichotomy or using the golden section strategy.
  • functions : Contains function utilities useful for the other main scripts.
  • data : Contains the two datasets presented in the paper.
  • plots and results_semi are results directories filled by executing semisupervised_tests.py.

About

A sparsified AutoEncoder to solve Semi-Supervised classification tasks

Topics

Resources

Stars

Watchers

Forks

Languages