SHADE

SHADE (Structure-preserving High-dimensional Analysis with Density-based Exploration) is a deep clustering algorithm that combines neural network-based dimensionality reduction with density-based clustering. It trains an autoencoder with reconstruction loss and a custom d_dc loss, followed by initial clustering by using the DCTree.

Installation

1. Clone the repository:

   git clone <repository-url>
   cd SHADE

2. Install the required dependencies:

   pip install numpy torch clustpy scikit-learn tqdm matplotlib

   Note: Ensure you have PyTorch installed compatible with your CUDA version if using GPU.

Usage

from shade import SHADE
import numpy as np

# Your data
X = np.random.rand(1000, 10)

# Initialize and fit SHADE
shade = SHADE(embedding_size=5, random_state=42)
shade.fit(X)

# Get cluster labels
labels = shade.labels_
print("Cluster labels:", labels)

Parameters

• batch_size: Size of the data batches (default: 500)
• embedding_size: Size of the embedding (default: 10)
• neural_network: Custom neural network (default: None, uses FeedforwardAutoencoder)
• optimizer_params: Optimizer parameters (default: {"lr": 1e-3})
• random_state: Random state for reproducibility (default: None)
• device: Device to run on (default: auto-detect)

Example

An example usage can be found in experiments/motivation/Motivation.ipynb. This notebook demonstrates SHADE on synthetic data, comparing it with other deep clustering methods.

To run the example:

1. Install Jupyter: pip install jupyter
2. Run: jupyter lab experiments/motivation/Motivation.ipynb

License

BSD 3-Clause License