rmems/Spikenaut-SNN-Telemetry

---

⚠️ Status: Ground Zero Rebuild (April 2026) This dataset has been intentionally purged and reset. After initial prototyping and large-scale data ingestion, the research has transitioned into a strict verification phase. To ensure the scientific integrity of the Spikenaut ecosystem, all previously uploaded data was deleted to make way for a verified, high-fidelity baseline.

Current Objective: Re-validating the hardware telemetry layer using Dynex mining as the ground-truth baseline.

🧠 Project Context This telemetry data serves as the "Sensory Input" for the Spikenaut Neuromorphic Architecture. Before infusing SNN logic into high-level models like OLMoE-7B, the underlying temporal signals must be absolute.

Research Focus: Neuromorphic High-Frequency Data Processing.

Methodology: "Measure twice, spike once."

Hardware Baseline: Pure Dynex mining telemetry (GPU/CPU/Efficiency).

Infrastructure: Developed on the Ship of Theseus workstation (Fedora 43).

📊 Dataset Structure (WIP) Once the verification process is complete, this dataset will contain high-resolution temporal features formatted for Liquid State Machine (LSM) and STDP training.

1. Raw Telemetry Data Timestamps: Microsecond-precision Unix epochs.

Compute Metrics: GPU NVML (Power, Temp, Clocks) and CPU k10temp/powercap.

Algorithm Efficiency: Hashrate fluctuations and mining pool volatility.

2. Spiking Features Poisson Encodings: Data translated into spike trains for SNN-native processing.

Neuromodulator Signals: Reward/Pain signals derived from efficiency vs. thermal overhead.

⚖️ License & Research Ethics This dataset is part of the open-source research initiative by Raul Montoya Cardenas.

License: GNU General Public License v3.0 (GPL-3.0)

Intent: Transparency and reproducibility in neuromorphic engineering.