uhg 0.1.12

Universal Hyperbolic Geometry Library for PyTorch

Universal Hyperbolic Geometry (UHG)

A PyTorch library for hyperbolic deep learning using pure UHG principles.

Features

• Pure projective geometry implementation
• No differential geometry or manifold concepts
• Cross-ratio preservation
• Projective transformations
• Graph neural networks
• Optimizers and samplers

Quick Start

import torch
import uhg

# Create points in projective space
x = torch.randn(10, 3)
y = torch.randn(10, 3)

# Initialize UHG
uhg_proj = uhg.ProjectiveUHG()

# Transform points
x_proj = uhg_proj.transform(x)
y_proj = uhg_proj.transform(y)

# Compute projective distance
dist = uhg_proj.proj_dist(x_proj, y_proj)

# Compute cross-ratio
cr = uhg_proj.cross_ratio(x_proj[0], x_proj[1], x_proj[2], x_proj[3])

Installation

pip install uhg

Documentation

For detailed documentation, visit uhg.readthedocs.io.