Kahlus / NeuroTwin research documentation#
Kahlus is a research codebase for leakage-safe neural translation benchmarks and NeuroTwin/NFC model development. These docs are organized like a scientific Python project: tutorials, methods, evidence, figures, API references, and paper-facing notes.
Current scientific posture
The repo should not claim a new brain foundation model or clinical digital twin. The defensible target is careful neural translation, missing-modality reconstruction, future-state forecasting, and subject/site generalization under leakage-safe splits.
What to read first#
Ridge EEG diagnostics for Amrith: the current visual packet for explaining why ridge regression can look strong.
Visual standards: what counts as reputable neuroscience visualization vs AI-generated slop.
Reputable neuroscience repo patterns: what MNE, MOABB, Braindecode, CEBRA, NeuroML, and Nilearn teach us.
Benchmark protocol: existing reviewer-facing benchmark rules.
Leakage-proof evaluation: split and audit logic.
Documentation map#
Research figures
Methods and evidence
Research design
- NeuroTwin Project State: NFC Pivot and Evidence-Gated Plan
- One-Paragraph Summary
- Architecture Hierarchy
- Track A and Track B
- What Old NeuroTwin Is Now
- What Pair-Operator Is Now
- What NFC Is Now
- What fNIRS Contributes
- What TurboQuant/TurboVec Contributes
- What Is Implemented
- What Is Scaffolded
- What Is Theory Only
- What Is Archived/Deferred
- Next Experiment Ladder
- Claim Boundaries
- Kill Criteria
- NeuroTwin Master Research State
- Current Repository State
- Track A: Reproducibility and Gates
- Track B: NFC Architecture
- Why Pair-Operator Was Demoted
- Why NFC Became the Primitive
- Equation Categories
- fNIRS Research Status
- TurboQuant/TurboVec Research Status
- Current Implementation Map
- Immediate Next Step
- Falsification Criteria
- A100 Run Order
- Claim Hygiene
- NeuroTwin NFC Mathematical Constitution
- Primitive
- Observation Operator
- Generative Model
- Controlled Dynamics
- Neural Field Dynamics
- Discretized Dynamics
- Low-Rank Pair Kernel
- fMRI Observation
- EEG and MEG Observation
- Spike, Calcium, and Behavior Observations
- Operator Learning Interpretation
- State-Space Interpretation
- Graph Calculus Interpretation
- Identifiability and Gauge Ambiguity
- Uncertainty and Calibration
- Synthetic Proving Ground
- Why Direct Fusion Is Not Enough
- Why Pair-Operator Is a Submodule
- Equation Ledger
- Falsification core (
src/neurotwin/falsification.py) - NeuroTwin Pair-Operator Historical Ablation Design
- Problem Statement
- Prior Art Table
- What Pair-Operator Adds
- Architecture Diagram
- State Definitions
- Low-Rank Controls
- Multi-Timescale Operator
- Stimulus Alignment Adapter
- Uncertainty Head
- Evidence Gate
- Non-Finite Safety
- Baselines
- Ablation Matrix
- Configs
- A100 Memory Estimates
- Exact Command Helper
- Failure Modes
- Paper Claim Criteria
- Kill Criteria
Runbooks and repo operation
External standards
Appendix and archived notes
- Chapman A100 Controlled Launch
- Chapman A100 Quickstart
- H100 Runbook
- License And Reuse
- RunPod A100 Rehearsal
- Ridge EEG diagnostic figures
- Repo Knowledge Graph
- Abstract Draft
- Paper Outline
- Related Work
- Code Architecture Map
- fNIRS Observation-Operator Notes
- Math Implementation Coverage
- NeuroTwin Technical Report
- NFC Architecture Decision Log
- NFC Falsification Criteria
- NeuroTwin NFC Implementation Plan
- Pair-Operator Readiness Report
- Repo Alignment Plan
- TurboQuant/TurboVec Retrieval Notes
- Coding Sprint 0–14 Days
- Kahlus Implementation Status
- Kahlus v2/v3/EM Sprint Ledger
- Kahlus-Affect / ResearchDock Roadmap
- Kahlus Current Mathematics and Computer Science Report
- Kahlus Neural-CASP general plan
- 0. Current repo state matters
- 1. Candidate moonshots, updated
- 2. Candidate A: Universal cross-modal neural-translation law
- 3. Candidate B: Gut-brain information channel, BCI↔GCI
- 4. Candidate C: predictive consciousness complexity measure
- 5. Candidate D: brain-as-critical-system order parameter
- 6. Candidate E: provable nuisance-invariant self-supervised objective
- 7. Candidate F: Neural-CASP benchmark
- 8. The new synthesis
- 9. Updated Kahlus moonshot ladder
- 10. What I would research/build next
- 11. Experiment 1: Criticality Residual
- [ \Delta RFS_{\text{crit}}
- 12. Experiment 2: BCI↔GCI scout
- [ \Delta RFS_{\text{EGG}}
- 13. Experiment 3: Predictive complexity index
- 14. Hard theorem-style insight
- [ I(Y;X,B,S,D)
- 15. The actual “revolutionary” claim if it works
- 16. Honest recommendation
- Bottom line
- Kahlus NV-1 Neurovisual Dataset Registry
- Kahlus-NeuroVisual Roadmap
- Kahlus NV-1 Neurovisual Symptom Ontology
- Kahlus Pathway Matrix
- Kahlus Root Technical Spec
- Kahlus-STF Benchmark Math Stop Note
- Kahlus-STF Public Dataset Review
- M4 Sleep-EDF Primary-Horizon Pre-Analysis Plan
- Overnight Log - 2026-07-03
- Overnight Proposals - 2026-07-03
- Overnight Result - 2026-07-03
- Kahlus Biomedical Execution Plan
- Kahlus Implementation State Ledger
- Kahlus v1 EEG Baseline Plan
- Kahlus v1 Few-Shot Adaptation Plan