interdim 0.4.0

Interactive Dimensionality Reduction, Clustering, and Visualization

InterDim

Interactive Dimensionality Reduction, Clustering, and Visualization

InterDim is a Python package for interactive exploration of latent data dimensions. It wraps existing tools for dimensionality reduction, clustering, and data visualization in a streamlined interface, allowing for quick and intuitive analysis of high-dimensional data.

Features

• Easy-to-use pipeline for dimensionality reduction, clustering, and visualization
• Interactive 3D scatter plots for exploring reduced data
• Support for various dimensionality reduction techniques (PCA, t-SNE, UMAP, etc.)
• Multiple clustering algorithms (K-means, DBSCAN, etc.)
• Customizable point visualizations for detailed data exploration

Installation

You can install from PyPI via pip (recommended):

pip install interdim

Or from source:

git clone https://github.com/MShinkle/interdim.git
cd interdim
pip install .

Quick Start

Here's a basic example using the Iris dataset:

from sklearn.datasets import load_iris
from interdim import InterDimAnalysis

iris = load_iris()
analysis = InterDimAnalysis(iris.data, true_labels=iris.target)
analysis.reduce(method='tsne', n_components=3)
analysis.cluster(method='kmeans', n_clusters=3)
analysis.show(n_components=3, point_visualization='bar')

This will reduce the Iris dataset to 3 dimensions using t-SNE, clusters the data using K-means, and displays an interactive 3D scatter plot with bar charts for each data point as you hover over them.

Demo Notebooks

For more in-depth examples and use cases, check out our demo notebooks:

1. Iris Species Analysis: Basic usage with the classic Iris dataset.
2. DNN Latent Space Exploration: Visualizing deep neural network activations.
3. LLM Token Analysis: Exploring language model token embeddings and layer activations.

Documentation

For detailed API documentation and advanced usage, visit our GitHub Pages.

Contributing

We welcome discussion and contributions!

License

InterDim is released under the BSD 3-Clause License. See the LICENSE file for details.

Contact

For questions and feedback, please open an issue on GitHub.