Hi everyone,
I've released FractalBrainOS v5.2 (Kubera Edition) — an open-source, self-learning neuromorphic brain engine implemented in C++17 without backpropagation.
Key Concepts & Architecture:
- Phase Dynamics: Hierarchical Kuramoto synchronization on self-similar triplet structures (3^L fractal scaling).
- Unsupervised Learning: Driven by Karl Friston's Free Energy Principle (FEP). Free energy F = |predicted - actual| phase error gates STDP plasticity.
- Multi-scale Memory: Clock oscillators (0.1s to 24h) provide temporal addressing, while standing waves in delay-lines store episodic patterns.
- Edge-Optimized: Quantized to int16 with precomputed sin/cos LUT tables (512KB footprint, runs efficiently on Raspberry Pi / mobile CPUs).
- Ethical Core: Hardcoded Kubera Dharma protocol (Ahimsa) ensuring harm-suppression via F-spikes.
- LLM Integration: Language models act purely as a Broca-like speech zone, translating thoughts without controlling decisions.
Source Code & Documentation:
👉 GitHub: https://github.com/NineNi999neNine/FractalBrainOS
The project is MIT licensed. I would love to hear feedback, critique, or ideas from researchers working on oscillatory neural networks and bio-inspired AI.
Top comments (0)