Code for ECMLPKDD 2019 Paper: A Framework for Deep Constrained Clustering - Algorithms and Advances

Installation

Step 1: Clone the Code from Github

git clone https://github.com/blueocean92/deep_constrained_clustering
cd deep_constrained_clustering

Step 2: Install Requirements

Python: see requirement.txt for complete list of used packages. We recommend doing a clean installation of requirements using virtualenv:

conda create -n testenv python=3.6
source activate testenv
pip install -r requirements.txt 

If you dont want to do the above clean installation via virtualenv, you could also directly install the requirements through:

pip install -r requirements.txt --no-index

PyTorch: Note that you need PyTorch. We used Version 1.0.0 If you use the above virtualenv, PyTorch will be automatically installed therein.

Running Constrained Clustering Experiments

While in deep_constrained_clustering folder:

Step 1: Download Pretrained Networks

sh download_model.sh

Step 2: Download Processed Reuters Data(optional, MNIST and Fashion is available in torchvision.datasets)

sh download_data.sh

cd experiments/

While in deep_constrained_clustering/experiments folder:

Step 3: Run Experimental Scripts to Reproduce Results

Option 1: Run Demo Pairwise Constraints Script

To run the pairwise constrained clustering using pre-trained weights (AE features, 6000 constraints), do:

python run_DCC_pairwise.py --data $DATA

For the --data flag which specifies the data set being used, the options are "MNIST", "Fashion" and "Reuters".

To run the pairwise without constrained clustering from raw features, do:

python run_DCC_pairwise.py --data $DATA --without_pretrain

To run the pairwise without KMeans initialization, do:

python run_DCC_pairwise.py --data $DATA --without_kmeans

To run the pairwise constrained clustering with noisy pairwise constraints do:

python run_DCC_pairwise.py --data $DATA --noisy $NOISE

For the --noisy flag which specifies the noisy degree, the option should be a positive float equal to the ratio of noisy constraints to ground truth constraints.

To save data for plotting, do:

python run_DCC_pairwise.py --data $DATA --plotting

This will save the experiment data for plotting in folders under ./plotting

To plot the results, do:

python ./plotting/plot_pairwise.py

Option 2: Run Demo Instance Constraints Script

To run the instance difficulty constrained clustering, do:

python run_DCC_instance.py --data $DATA

Option 3: Run Demo Triplets Constraints Script

To run the triplets constrained clustering (6000 constraints), do:

python run_DCC_triplets.py --data $DATA

Option 4: Run Demo Global Constraints Script

To run the global size constrained clustering, do:

python run_DCC_global.py --data $DATA

Option 5: Run Demo Improved DEC Script

To run the baseline Improved DEC, do:

python run_improved_DEC.py --data $DATA