SC-NeuroCore v3.7.0 — Polymorphic Engine

SC-NeuroCore v3.7.0 — "Polymorphic Engine"

Three new modules turn the SNN simulator into a general-purpose stochastic processor.

Module A — SCPN Fusion Engine (PetriNetEngine)

• Stochastic Colored Petri Net simulator via matrix algebra
• Wraps two DenseLayer instances (W_in, W_out) for place→transition→place firing
• Configurable thresholds, fractional token arithmetic, step counting

Module B — HDC/VSA Kernel (HDCVector + BitStreamTensor)

• Hyper-Dimensional Computing with 10,000-bit binary vectors
• Bind (* / XOR) — self-inverse concept association
• Bundle (+ / majority vote) — set superposition
• Permute (cyclic rotation) — sequence/position encoding
• Similarity (1 − normalised Hamming distance) — relatedness measure
• SIMD-accelerated fused XOR+popcount (AVX-512 / AVX2 / NEON / portable)

Module C — Fault-Tolerant Binary Streams

• Safety-critical Boolean logic via stochastic redundancy
• 1024-bit encoded Booleans tolerate up to ~45% random bit-flip noise
• Bitwise AND/OR/NOT on packed words, majority-vote decoding

New Python API

from sc_neurocore_engine import BitStreamTensor, HDCVector, PetriNetEngine

# HDC symbolic reasoning
france = HDCVector(10000, seed=1)
paris  = HDCVector(10000, seed=2)
bound  = france * paris          # XOR bind
print(bound.similarity(france))  # ~0.50 (quasi-orthogonal)

# Petri Net simulation
import numpy as np
engine = PetriNetEngine({
    'w_in': np.eye(3), 'w_out': np.eye(3),
    'thresholds': np.array([0.5, 0.5, 0.5]),
    'marking': np.array([1.0, 0.0, 0.0])
}, length=2048, seed=42)
engine.step(seed=100)

Infrastructure

- 12 pre-built wheels: Linux / macOS / Windows × Python 3.9–3.12
- Jupyter notebooks: 01_hdc_symbolic_query.ipynb, 02_fault_tolerant_logic.ipynb
- Rustdoc: https://anulum.github.io/sc-neurocore/
- GitHub Actions: wheel CI + rustdoc deployment + PyPI trusted publishing

Full Changelog

https://github.com/anulum/sc-neurocore/compare/v3.6.0...v3.7.0