Tensor Earthquake Classifier

What it does

Takes the 9,700-dimensional Earth State Tensor as input, projects onto a 50-dim eigenmode + non-zero-feature subspace, and predicts whether a global M>=5.5 earthquake will occur within the next 3 days. Trained on real USGS outcome labels for each historical tensor snapshot, not on synthetic distributions.

Physics basis

The tensor fuses seismic dv/v, Schumann resonance, telluric currents, GNSS strain, IGETS superconducting gravimeters, cosmic-ray muon flux, and 150+ other signals known from the peer-reviewed literature to respond to crustal stress loading. The classifier learns which combinations discriminate pre-event from quiet periods -- not from hand-tuned features but from the real outcome history.

When it fires

Every 15 minutes, after each tensor snapshot. High-confidence predictions are logged to the ledger as tensor_eq_per_region or tensor_eq_global.

What the customer receives

• Probability estimate every 15 min for global M>=5.5 in next 3 days
• Per-region variants (54 regions) at lower magnitude threshold
• SHA-256 signed ledger entries on every high-confidence fire
• Weekly retraining log showing AUC + Brier drift

Operational numbers (live)

Engagement paths

Per-query, subscription, territorial-exclusive, and royalty-on-find structures are all available. Specific commercial terms are scoped after a technical-fit conversation.

Honest caveats

• 67 snapshots with real outcomes is small; temporal correlation between neighboring snapshots means effective independent sample count is lower.
• AUC will shift as snapshots accumulate; per-region classifiers need 6-12 months of tensor data before some sparse regions become trainable.
• Confidence gate is set loosely (>=0.40) because tensor probabilities are calibrated via isotonic CalibratedClassifierCV -- not the raw logistic output.