KNRscore Documentation

Overview

KNRscore is a powerful Python package for quantitative comparison of high-dimensional embeddings using a scale-dependent similarity measure. It enables researchers and data scientists to:

Compare different dimensionality reduction techniques (PCA, t-SNE, UMAP, etc.)
Quantify local similarities between embeddings
Evaluate the preservation of neighborhood structures
Visualize comparison results with intuitive plots

Key Features

Scale-dependent Analysis: Compare embeddings at different neighborhood scales
Flexible Input: Works with any embedding or high-dimensional data
Intuitive Visualization: Generate clear comparison plots
Easy Integration: Simple usage with comprehensive documentation

Tip

For a detailed explanation of the methodology, check out our Medium blog post on quantitative comparisons between dimensionality reduction techniques.

Note

Your ❤️ is important to keep maintaining this package. You can support in various ways, have a look at the sponser page. Report bugs, issues and feature extensions at github page.

Quick Start

pip install KNRscore

import KNRscore as knrs
from sklearn.decomposition import PCA
from sklearn.manifold import TSNE

# Load example data
X, y = knrs.import_example(data='digits')

# Create embeddings
pca = PCA(n_components=2).fit_transform(X)
tsne = TSNE(n_components=2).fit_transform(X)

# Compare embeddings
scores = knrs.compare(pca, tsne)

# Visualize results
fig, ax = knrs.plot(scores, xlabel='PCA', ylabel='tSNE')

Note

Support the Project: Your support helps maintain and improve this package. Consider:

Starring the repository on GitHub
Reporting issues and suggesting features
Contributing code or documentation
Supporting through GitHub Sponsors

Visit our sponsor page for more ways to contribute.

Documentation Contents

Getting Started

Examples

Examples

Reference