Skip to content

Mat-Tom-Son/memory-resonance

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Memory-Resonance Condition: Cross-Domain Control for Colored-Noise Systems

Paper License Status

This repository contains the simulation code, data, and analysis for our paper on the Memory-Resonance Condition (MRC): a cross-domain design principle showing that systems across classical and quantum domains exhibit optimal performance when environmental memory synchronizes with internal dynamics (Θ ≡ ω_fast τ_B ≈ 1).


Quick Links

  • Paper: mrl_synthesis_paper/main.pdf (latest build)
  • Data: Core CSVs: results/theta_sweep_today.csv (Class S), results/classical_parametric_mod03.csv (Class C), results/quantum_nonlin/kerr02_tol001.csv (Class M), results/quantum_nonlin/kerr00_tol001.csv (quantum surrogate), results/quantum_eqheat_sweep_for_figB.csv (quantum null)
    • Supplementary (adaptive confirmations): results/adaptive/classC_fraction_attempt.csv (Class C re-run), results/adaptive/kerr08_det12_theta_grid.csv (Class M fast grid). These confirm the same diagnostics with tighter sampling; no model changes, only runtime/sampling refinements.
  • Figures: figures/ (scripts + PDFs for all panels)

What is the MRC?

Memory-Resonance Condition: Systems often perform best when their environment's correlation time (τ_B) matches their fastest internal rhythm (1/ω_fast):

Θ ≡ ω_fast τ_B ≈ 1

Key Insight: The observable (a shallow optimum near Θ≈1) recurs across substrates—from stochastic resonance in neural circuits to noise-assisted quantum transport—but the mechanism varies:

  • Class S (Spectral): Frequency-domain overlap in near-linear systems
  • Class C (Coherent): Weak nonlinearity reweights spectra
  • Class M (Memory): Time-nonlocal dynamical kernels

Our Contribution: We formalize this cross-domain pattern, provide operational diagnostics to classify mechanism, and apply them to a minimal hierarchy—confirming the classical mechanism while documenting where the quantum diagnostic returns a null.


Repository Structure

.
├── mrl_synthesis_paper/          # Paper source (LaTeX + PDF)
│   ├── main.tex                  # Main paper source
│   ├── main.pdf                  # Compiled PDF (390 KB, 11 pages)
│   └── references.bib            # Bibliography
│
├── figures/                      # Figure generation scripts
│   ├── make_figA.py             # Classical Class S (OU ≈ surrogate)
│   ├── make_figB.py             # Quantum linear equal-carrier (null)
│   ├── make_figC.py             # Robustness across metrics
│   ├── make_figD_parametric.py  # Classical Class C (parametric modulation)
│   ├── make_figE_collapse.py    # Cross-domain collapse onto Θ
│   ├── make_figF_quantum_nonlin.py # Quantum Class M (detune + Kerr)
│   ├── figA_classical.pdf
│   ├── figB_equal_carrier.pdf
│   ├── figC_robustness.pdf
│   ├── figD_parametric_classC.pdf
│   ├── figE_collapse.pdf
│   └── figF_quantum_nonlin.pdf
│
├── results/                      # Simulation data and manifests
│   ├── theta_sweep_today.csv    # Consolidated dataset (301 KB)
│   └── production_archive/      # QA logs, manifests, config hash
│
├── src/                         # Core simulation code
│   ├── quantum_models.py        # Quantum hierarchy (Lindblad + pseudomode)
│   ├── hierarchical_analysis.py # Classical analysis (OU bath, controls)
│   └── ...                      # Supporting modules
│
├── tests/                       # Validation suite
│   └── test_quantum_hierarchy.py
│
├── PAPER_STATUS.md              # Current paper status
├── SUBMISSION_READY.md          # Pre-submission checklist
└── README.md                    # This file

Quick Start

1. Setup Environment

# Clone repository
git clone https://github.com/Mat-Tom-Son/memory-resonance.git
cd memory-resonance

# Create virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install dependencies
pip install --upgrade pip
pip install -r requirements.txt

2. Reproduce Paper Figures

# Regenerate all figures from consolidated data
cd figures
PYTHONPATH=.. python make_figA.py              # Classical Class S (~5 sec)
PYTHONPATH=.. python make_figD_parametric.py   # Classical Class C (~8 sec)
PYTHONPATH=.. python make_figF_quantum_nonlin.py # Quantum Class M (~8 sec)
PYTHONPATH=.. python make_figB.py              # Quantum null (~5 sec)
PYTHONPATH=.. python make_figE_collapse.py     # Cross-domain collapse (~3 sec)
PYTHONPATH=.. python make_figC.py              # Robustness (~5 sec)

Output: figA_classical.pdf, figD_parametric_classC.pdf, figF_quantum_nonlin.pdf, figB_equal_carrier.pdf, figE_collapse.pdf, figC_robustness.pdf

  • Supplementary adaptive figures: figures/figD_parametric_adaptive.pdf, figures/figF_quantum_nonlin_adaptive.pdf

3. Run Tests

# Validate simulation code
pytest tests/test_quantum_hierarchy.py -v

# Quick diagnostic (10 seconds)
python quick_diagnostic.py

For Peer Reviewers

To quickly verify the paper's claims:

  1. Check the data:
    • head results/theta_sweep_today.csv (Class S)
    • head results/classical_parametric_mod03.csv (Class C) plus head results/classical_parametric_mod00.csv (no-modulation control)
    • head results/quantum_nonlin/kerr02_tol001.csv (Class M), head results/quantum_nonlin/kerr00_tol001.csv (linear detuned control), head results/quantum_nonlin/kerr02_tol001_neg.csv (negative detuning)
    • head results/quantum_eqheat_sweep_for_figB.csv (quantum null)
  2. Regenerate figures: cd figures and run PYTHONPATH=.. python make_figA.py, PYTHONPATH=.. python make_figD_parametric.py, PYTHONPATH=.. python make_figF_quantum_nonlin.py, PYTHONPATH=.. python make_figB.py, PYTHONPATH=.. python make_figE_collapse.py, PYTHONPATH=.. python make_figC.py
  3. Run validation suite: pytest tests/test_quantum_hierarchy.py -v (~20 sec)
  4. Config hash verification: All figures embed c7dc5aa1 - grep for it in the PDF or check figure captions

Expected verification time: < 5 minutes


What's Novel Here?

We do not claim novelty of the Θ≈1 phenomenon itself (it appears across stochastic resonance, coherence resonance, and noise-assisted transport).

Our contribution:

  • Pattern recognition: Formalizing the cross-domain recurrence as a unified Memory-Resonance Condition
  • Mechanism taxonomy: Classes S/C/M with operational diagnostics to distinguish spectral overlap, coherent modulation, and memory backaction
  • Falsifiable tests in practice: Class S replication (OU ≈ surrogate), a parametric modulation run that breaks the PSD gate (Class C), and a detuned Kerr equal-carrier sweep that yields a quantum enhancement (Class M) alongside the linear null
  • Rigorous validation + boundaries: Pre-registered gates reported for all sweeps (PSD-NRMSE, |d_z|, |�95J|/J*), highlighting where the diagnostics pass, fail, or mark the edge of applicability
  • Actionable design rule: Boxed Design Card with step-by-step guidance and shared CSVs for immediate reuse

Key Results

Results Summary

Pillar Mechanism Gate Threshold Observed Status
Classical Spectral overlap (Class S) PSD-NRMSE < 0.03 0.006–0.007 ✓ Pass
Classical Coherent modulation (Class C) PSD-NRMSE 1.08–2.13 ✗ Surrogate fails
Classical Coherent modulation (Class C) |d_z| ≈0.20 ✗ Surrogate fails
Quantum Memory backaction (Class M) |ΔJ|/J* ≤ 0.02 ≤ 1e-3 ✓ Pass (peak R_env ≈ 1.11 at Θ=0.90; adaptive)
Quantum Linear boundary |ΔJ|/J* ≤ 0.02 < 0.02 ✓ Pass (R_env ≈ 1.00)
All Robustness Metric consistency - Baseband ≈ Narrowband ✓ Pass

Classical Pillar (Class S - Spectral Overlap)

Finding: Ornstein-Uhlenbeck (OU) noise and PSD-matched surrogates are practically equivalent.

Gates: PSD-NRMSE < 0.03, |d_z| < 0.30 ✓

Interpretation: Classical Θ-dependence arises from spectral overlap (Wiener-Khinchin), consistent with stochastic resonance literature.

Classical Coherent Modulation (Class C)

Finding: A weakly modulated coupling yields a Θ-band optimum ($R_{\mathrm{env}}\approx 1.31$ at Θ ≈ 1.1) even though the PSD-matched surrogate stays near unity; PSD-NRMSE $&gt;1$ violates the spectral-overlap gate.

Quantum Probes (Class M)

  • Detuned Kerr sweep: Equal-carrier calibration at the operating amplitude ($|�ΔJ|/J^*|�≤�10^{-3}$) yields a clear peak in the adaptive fast scan at Θ = 0.90 (R_env ≈ 1.11); auxiliary metrics agree and the fast-mode occupancy remains ≈2×10⁻².
  • Detuned linear surrogate: With Kerr set to zero the equal-carrier curve stays flat (R_env ≤ 1.06), demonstrating that mild nonlinearity is essential for the enhancement.
  • Linear-Gaussian null: Without detuning the original hierarchy returns R_env ≈ 1 for all Θ, marking the boundary of the minimal hierarchy.

Robustness

Finding: Baseband and narrowband metrics agree in ordering across Θ.

Interpretation: MRC is a system-environment property, not a metric-specific artifact.


Reproduce from Scratch

Classical Analysis (Long Run)

# Full OU calibration + surrogate comparison + Θ sweep
python hierarchical_analysis.py  # ~20 minutes, 25 Θ points × 16 seeds

# Outputs:
# - experiment1_equal_variance.png
# - experiment2_sweep_equal_variance.png
# - Terminal: Bootstrap CI on Θ peak

# Class C parametric modulation sweep
python classical_parametric_sweep.py --output results/classical_parametric_mod03.csv --mod-amp 0.3 --n-seeds 6
PYTHONPATH=. python figures/make_figD_parametric.py

# (Optional) no-modulation control for collapse/checks
python classical_parametric_sweep.py --output results/classical_parametric_mod00.csv --mod-amp 0.0 --n-seeds 6

Quantum Hierarchy

# Stage 1: Markovian baseline
python stage1_markovian.py  # ~30 sec

# Stage 2: Pseudomode comparison
python stage2_pseudomode.py  # ~1 min

# Stage 3: τ_B parameter sweep
python stage3_parameter_sweep.py  # ~5 min

# Or run complete pipeline:
python quantum_hierarchy.py  # ~6 min

# Class M (detuned + Kerr) vs surrogate
python quantum_nonlin_sweep.py --output results/quantum_nonlin/kerr02_tol001.csv --detune-frac 0.1 --kerr-fast 0.02 --equal-carrier-tol 0.001 --theta-list 0.6 0.8 1.0 1.1 1.2 1.3 1.4
python quantum_nonlin_sweep.py --output results/quantum_nonlin/kerr00_tol001.csv --detune-frac 0.1 --kerr-fast 0.0 --equal-carrier-tol 0.001 --theta-list 0.6 0.8 1.0 1.1 1.2 1.3 1.4
python quantum_nonlin_sweep.py --output results/quantum_nonlin/kerr02_tol001_neg.csv --detune-frac -0.1 --kerr-fast 0.02 --equal-carrier-tol 0.001 --theta-list 0.6 0.8 1.0 1.1 1.2 1.3 1.4
PYTHONPATH=. python figures/make_figF_quantum_nonlin.py

Paper Highlights

Design Card (Actionable Guidance)

The paper includes a boxed Design Card with:

  1. Rule: Target Θ in the MR band [0.7, 1.4]
  2. How to estimate ω_fast: 3 cases (linear, nonlinear, driven)
  3. How to choose τ_B: Use τ_B^(eff) (observable-effective timescale)
  4. Diagnostics: Tests to classify Classes S/C/M
  5. Failure modes: When MRC may not apply
  6. Optional controller: Two-point dither for τ_B adaptation

Synthesis Map (Cross-Domain Evidence)

The paper synthesizes evidence from:

  • Stochastic resonance (Mondal et al. 2018, Gammaitoni et al.)
  • Coherence resonance (Brugioni et al. 2005, Pikovsky & Kurths)
  • Quantum transport (Moreira et al. 2020, Plenio & Huelga)
  • Photosynthesis (Uchiyama et al. 2017)
  • Neural detection (Duan et al. 2014)
  • Energy harvesting (Romero-Bastida & López 2020)

Data Availability

All simulation code, raw data, configuration manifests, and figure-generation scripts are in this repository:

  • Consolidated dataset: results/theta_sweep_today.csv (301 KB)
  • Config hash: c7dc5aa1 (production run identifier)
  • Reproduction scripts: figures/make_fig*.py
  • QA manifests: results/production_archive/

Reproducibility: All figures can be regenerated from source data. Gates, seeds, and quality checks are documented.


Advanced Usage

Custom Parameter Sweeps

# Modify stage3_parameter_sweep.py
tau_B_values = np.linspace(0.1, 3.0, 50)  # Denser Θ grid
# Then run: python stage3_parameter_sweep.py

Custom Observables

# Add to quantum_models.py
def compute_entanglement_entropy(rho, subsystem_dims):
    # Your implementation
    return S_ent

# Use in stage scripts to plot new metrics

Performance Tuning

# Estimate runtime for large Hilbert spaces
python benchmark.py

# For classical runs, reduce n_seeds or T for exploration:
# In hierarchical_analysis.py:
n_seeds = 8       # Default: 16
T = 100.0         # Default: 200.0

Citation

If you use this code or data in your research, please cite:

@article{Thompson2025MRC,
  title={Memory-Resonance Condition: A Cross-Domain Control Principle for Colored-Noise Systems},
  author={Thompson, Mat},
  journal={Preprint},
  year={2025},
  note={In preparation; see repository for latest PDF}
}

(We will update with the arXiv identifier upon posting.)


FAQ

Q: Is this the same as stochastic resonance?

A: Similar observable (Θ≈1 optimum), but we recognize it as a broader cross-domain pattern. Stochastic resonance is typically Class S (spectral overlap). The MRC framework also covers Class C (coherent modulation) and Class M (memory backaction).

Q: Can I use this for my system?

A: Yes! Check if your system has: (1) timescale separation between fast and slow dynamics, (2) a tunable environmental correlation time τ_B, and (3) a measurable slow-band observable. See the Design Card in the paper (after Methods section) for step-by-step guidance.

Q: What if I have multiple ω_fast peaks?

A: See "Failure modes" in the Design Card - multimode ambiguity is a known limitation. The MRC applies best when there's a single dominant transduction frequency. For multi-peak systems, you may need controller synthesis instead of passive tuning.

Q: How do I know which mechanism class (S/C/M) applies to my system?

A: Run the diagnostics:

  • Class S test: Does a PSD-matched surrogate reproduce the Θ≈1 optimum? (Pass → Class S)
  • Class M test: Does the optimum persist under equal-carrier conditions (fixed spectral weight)? (Yes → Class M)
  • Class C: If both tests fail, inspect for weak nonlinearity or coherent modulation effects

Q: What if my τ_B isn't tunable?

A: See the "Optional controller" in the Design Card. You can use a two-point dither or sample τ_B from the MR band [0.7, 1.4] × (1/ω_fast) to hedge against model mismatch.

Q: Why pre-register gates instead of using p-values?

A: Practical equivalence gates (PSD-NRMSE < 0.03, |d_z| < 0.30) are more meaningful than statistical significance for equivalence testing. We report p-values for transparency (in supplement), but gates are the primary acceptance criterion.


Contact & Contributing


License

This project is licensed under the MIT License - see LICENSE file for details.


Acknowledgments

  • QuTiP for quantum simulation framework
  • All cited authors for cross-domain evidence synthesis
  • Reviewers and collaborators for feedback

Additional Documentation

  • Paper Status: PAPER_STATUS.md - Current version, stats, checklist
  • Submission Ready: SUBMISSION_READY.md - Final checks, talking points
  • Session Notes: .claude_session_notes/ - Detailed revision history (hidden)

Status: Preprint ready for submission (2025-10-12)

Config hash: c7dc5aa1 | PDF: 390 KB, 11 pages | Data: 301 KB CSV

About

Simulations and analysis for the memory–resonance condition (Θ ≈ 1)

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors