lmflrb / gjrd

The generalized Jensen-Renyi's divergence (GJRD), serving as a robust tool for estimating the divergence between multiple distributions, is introduced to handle machine learning problems involving data from multiple distributions or sources. This work specifies GJRD-based deep clustering as a case study.

Generalized Jensen-Renyi's Divergence (GJRD)

Introduction

The generalized Jensen-Renyi's divergence (GJRD), serving as a robust tool for estimating the divergence between multiple distributions, is introduced to handle machine learning problems involving data from multiple distributions or sources. This work specifies GJRD-based deep clustering as a case study.

Code utilization

1. Install Required Packages:

   First, install the necessary packages as specified in requirement.txt:

   pip install -r requirement.txt

2. Train the Model: You can train the required model by running run_dataset_loss.py. Users can modify the code to decide which datasets or network configurations to run. Some basic settings are written in the configs.yaml file:

   python run_dataset_loss.py

3. python run_dataset_loss.py Users can also directly obtain the results of our stored model by running eval_pretrained_model.py:

python eval_pretrained_model.py

Feel free to modify the configurations and experiment with different settings to fit your needs.