Show HN: AI latent space with overlapping manifolds

Category: library

Tags: latent-space, topological-loss, pytorch

Score: 4.3/10 (Innovation: 5, Technical: 6, Documentation: 4, Utility: 2)

Egregore is a PyTorch-based differentiable manifold morphing engine that dynamically maps latent space data to sphere or torus manifolds using topological loss and guardrails. The project explores novel combinations of geometric topology and entropy concepts for machine learning, but is very niche and lacks broad utility or clear real-world application.

Target audience: ai-ml researchers

Repository: https://github.com/PJHkorea/Egregore/blob/main/integrated_egregore_core_test_v6_4.py · Python · GPL-3.0 · 3 stars

View on Hacker News