synos-neuromorphic-computing — Spiking Neural Network Framework

tags: [ai consciousness cortex]

synos-neuromorphic-computing — Spiking Neural Network Framework

Classification: PUBLIC
Crate: fruit/crates/synos-neuromorphic-computing/
Milestone: v41+
License: MIT OR Apache-2.0

What It Is

synos-neuromorphic-computing provides the full neuromorphic computing framework for Syn_OS. It implements spiking neural networks (SNNs) with multiple neuron models, synaptic plasticity, and homeostatic scaling. This is the computation engine behind ALFRED’s Neuromorphic consciousness path.

Architecture

Neuron Models

Model	Description
LIF	Leaky Integrate-and-Fire
Izhikevich	Izhikevich model
HodgkinHuxley	Hodgkin-Huxley model
AdEx	Adaptive Exponential Integrate-and-Fire
SRM	Spike Response Model

Synapse

Synapse represents a connection between neurons with:

• Weight (synaptic strength)
• Plasticity rules (STDP)
• Delay (conduction delay)

Key Types

• ConsciousnessState — shared state type from synos-consciousness-types
• NeuronModel — enum of supported neuron models
• Synapse — synaptic connection with plasticity

How It’s Wired

Integration Points

1. ALFRED cortex — CortexAdapter delegates to this crate for SNN inference
2. synos-consciousness-types — shares ConsciousnessState type
3. synos-cortex-q — alternative backend (MPS tensor-network) for QuantumInspired path
4. Criterion benchmarks — performance regression suite included

Future Ideas

1. Online learning — real-time STDP during inference
2. Homeostatic scaling — automatic neuron excitability regulation
3. Multi-compartment neurons — spatially extended neuron models
4. Hardware acceleration — GPU/TPU kernel for large-scale SNN simulation