README.md

metadata

title: μ-Net Training Lab
emoji: 🧬
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: 5.23.0
app_file: app.py
pinned: true
license: mit
short_description: Train neural networks grounded in the Eigenverse

🧬 μ-Net Training Lab

Train neural networks whose architecture IS the Eigenverse.

• 8 layers (μ⁸ = 1)
• μ^k phase-modulated activations (135° rotation per layer)
• Coherence loss regularization C(r) = 2r/(1+r²)
• Silver-gated skip connections (η = 1/√2)

552 Lean theorems → network architecture → trained weights.

Eigenverse · COINjecture