Quick Start Guide

Installation

To install TorchDR, run in the terminal:

pip install torchdr

When to use TorchDR

  • For leveraging the power of GPU acceleration for faster computations. All the modules in TorchDR are designed to work seamlessly on the GPU by setting device = 'cuda'.

  • For comparing different dimensionality reduction methods in a fair and reproducible way. TorchDR maximizes code sharing across various methods, ensuring a fair benchmarking that emphasizes core differences.

  • For developing new dimensionality reduction approaches. TorchDR provides a modular and extensible framework that allows you to focus on the core ideas of your method, while the rest of the pipeline is taken care of.

Examples using AffinityMatcher:

PCA via SVD and via AffinityMatcher

PCA via SVD and via AffinityMatcher

TSNE embedding of the swiss roll dataset

TSNE embedding of the swiss roll dataset

Neighbor Embedding on genomics & equivalent affinity matcher formulation

Neighbor Embedding on genomics & equivalent affinity matcher formulation