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.
-
Install Required Packages:
First, install the necessary packages as specified in
requirement.txt
:pip install -r requirement.txt
-
Train the Model: You can train the required model by running
run_dataset_loss.py
. Users can modify the code to decide whichdatasets
ornetwork
configurations to run. Some basic settings are written in theconfigs.yaml
file:python run_dataset_loss.py
-
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.