Clustering

Implementation of unsupervised clustering algorithms from scratch in different machine learning frameworks. The goal is to demonstrate the similarities and differences of the frameworks. If you have an idea to improve an implementation (e.g., a more elegant or faster solution) or would like to implement a different algorithm/framework, please feel free to contribute.

Clustering algorithms

• K-Means
• Mean shift

Machine learning frameworks

• NumPy
• PyTorch
• TensorFlow (Eager and Graph Mode)
• JAX

Algorithms

Algorithm	Framework
K-Means	NumPy	kmeans_np.py
	PyTorch	kmeans_pt.py
	TensorFlow 2 (Eager)	kmeans_tf_eager.py
	TensorFlow 2 (Graph)	kmeans_tf.py
	JAX	kmeans_jax.py
Mean shift	NumPy	meanshift_np.py
	PyTorch	meanshift_pt.py
	JAX	meanshift_jax.py
	TensorFlow 2 (Eager)	meanshift_tf_eager.py
	TensorFlow 2 (Graph)	meanshift_tf.py

Usage

Please follow the installation guide.

You can simply run the following command to execute a mean shift clustering on aggregation with JAX

python main.py 

To select different algorithms, datasets, or frameworks, r

The algorithm, dataset, and framework can be selected via command like options, set

• algorithm to kmeans or meanshift
• dataset to aggregation, jain, moons, s4, or meanshift
• framework to numpy, pytorch, jax, tensorflow_eager, or tensorflow

For example

python main.py algorithm=kmeans dataset=moons framework=pytorch

For all options, please see configs/base.yaml.

For timing, set time=true

python main.py time=true

Plot result

python main.py plot=true

Plot gif

python main.py plot_gif=true

Installation

Clone repository

git clone git@github.com:creinders/ClusteringAlgorithmsFromScratch.git
cd ClusteringAlgorithmsFromScratch

Install anaconda environment and dependencies

conda create -n clustering python=3.9
conda activate clustering

# Install PyTorch (follow https://pytorch.org/get-started)
conda install pytorch -c pytorch

# Install TensorFlow (follow https://www.tensorflow.org/install/pip)
pip install tensorflow

# Install JAX (follow https://github.com/google/jax#installation)
pip install --upgrade "jax[cuda]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

pip install -r requirements.txt

If you want to use the datasets aggregation, jain, or s4, please download the data

./download_datasets.sh