Golden Studio 🌀

Evolutionary optimization framework for vibroacoustic therapy plates using physics-informed multi-objective algorithms

---

🎯 What is this?

Golden Studio is a novel hybrid evolutionary framework that combines:

• Multi-objective NSGA-II for Pareto optimization
• Curriculum learning via staged gene activation (SEED→SPROUT→GROW→BLOOM)
• Neural memory (RDNN) for trajectory prediction across optimization runs
• Long-term knowledge distillation (LTM) for cross-run learning
• Physics-informed fitness based on FEM modal analysis

The goal: optimize DML (Distributed Mode Loudspeaker) plates for vibroacoustic therapy, where the human body is the "string to tune".

---

🏗️ Architecture

binaural_golden/src/core/
├── OPTIMIZATION LAYER
│   ├── unified_optimizer.py    # Strategy pattern (GENETIC/NSGA2/SIMP/HYBRID)
│   ├── agnostic_evolution.py   # Domain-agnostic interfaces
│   ├── evolution_pipeline.py   # Full pipeline: Pokayoke → Physics → RDNN → LTM
│   └── exciter_gene.py         # GenePhase curriculum learning
│
├── PHYSICS LAYER
│   ├── plate_physics.py        # Analytical modal analysis
│   ├── jax_plate_fem.py        # JAX-accelerated FEM (autodiff)
│   └── materials.py            # Spruce, Baltic birch, MDF...
│
├── FITNESS LAYER
│   ├── fitness.py              # Multi-objective evaluator
│   ├── scorers/                # Modular scoring (ear_uniformity, spine_coupling...)
│   └── structural_analysis.py  # ABH detection, peninsula analysis
│
├── MEMORY LAYER
│   ├── evolution_memory.py     # Short-term (ring buffer) + Long-term
│   └── rdnn_memory.py          # LSTM/GRU for trajectory prediction
│
└── EXPORT LAYER
    ├── stl_export.py           # 3D STL for CNC
    └── virtual_cnc.py          # G-code generation

---

🔬 Research Foundation

Built on 70+ distilled research papers covering:

Domain	Papers	Key References
Multi-exciter DML	8	Lu 2012, Bai & Liu 2004
Acoustic Black Holes	7	Krylov 2014, Deng 2019
Lutherie	12	Schleske 2002, Woodhouse 2014
Human Body Resonance	8	Griffin 1990 (spine 10-12Hz, chest 50-60Hz)
Vibroacoustic Therapy	7	Skille 1989 (30-120Hz), Boyd-Brewer 2003

Full bibliography: binaural_golden/docs/research/vibroacoustic_references.bib

---

✨ Key Features

🧬 Curriculum Learning (GenePhase)

class GenePhase(Enum):
    SEED = 0    # Position-only mutations (exploration)
    SPROUT = 1  # + diameter mutations
    GROW = 2    # + exciter count mutations
    BLOOM = 3   # Full mutation (position + diameter + count + emission)

Inspired by NEAT (Stanley 2002) and curriculum learning (Bengio 2009).

🧠 Neural Memory (RDNN)

# LSTM/GRU maintains hidden state across optimization runs
memory = RDNNMemory(architecture=RDNNArchitecture.GRU, hidden_size=64)
memory.observe(fitness_trajectory, physics_features)
suggestions = memory.suggest()  # Adaptive mutation rate, search direction

📊 Multi-Objective Pareto

objectives = ObjectiveVector(
    spine_flatness=0.85,      # 20-200Hz response at spine
    ear_lr_uniformity=0.97,   # L/R balance for binaural
    structural_safety=1.0,    # Deflection < 10mm
    abh_benefit=0.72,         # Acoustic Black Hole energy focusing
)
# NSGA-II finds Pareto-optimal trade-offs

🎻 Physics-Informed Fitness

• Modal analysis via analytical Kirchhoff or JAX-accelerated FEM
• Zone-specific response: spine (tactile 20-200Hz) vs head (audio 50-8kHz)
• ABH peninsula detection for energy focusing (Krylov 2014)

---

🚀 Quick Start

Installation

# Clone the repository
git clone https://github.com/Alemusica/Rememberance.git
cd Rememberance/binaural_golden

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

# Install dependencies
pip install -r requirements.txt

Run the GUI

python src/golden_studio.py

Run Optimization (CLI)

from src.core.person import Person
from src.core.unified_optimizer import OptimizationStrategy
from src.core.plate_unified import create_plate_optimization_system

# Create person model (the "string to tune")
person = Person(height_m=1.80, weight_kg=75)

# Create optimizer
optimizer, config = create_plate_optimization_system(
    person,
    strategy=OptimizationStrategy.NSGA2,
    use_memory=True
)

# Run evolution
result = optimizer.optimize(config)
print(f"Best fitness: {result.best_fitness:.4f}")
print(f"Pareto front size: {len(result.pareto_front)}")

---

📐 The Mathematics

Golden Ratio Constants

φ (Phi)           = 1.618033988749895...  (Golden Ratio)
φ conjugate       = 0.618033988749895...  (1/φ = φ-1)
Golden angle      = 2π/φ² ≈ 137.5°

Zone Frequency Targets

Zone	Frequency Range	Purpose
Spine	20-200 Hz	Tactile vibration therapy
Head/Ears	50-8000 Hz	Binaural audio reproduction
Chest	50-60 Hz	Resonance coupling

ABH (Acoustic Black Hole) Profile

h(x) = h₀ × (x / x_abh)^m   # m ≥ 2 for wave trapping

---

🧪 Tests

cd binaural_golden

# Run all tests
pytest tests/ -v

# Run physics validation tests
pytest tests/test_physics_validation.py -v  # 11/11 pass

# Run evolution pipeline tests
pytest tests/test_evolution_pipeline.py -v

---

🤝 Contributing

Contributions welcome! See CONTRIBUTING.md for guidelines.

Areas of interest:

• Ray parallelization for distributed evaluation
• Quality-Diversity (MAP-Elites) for zone-specific exploration
• Additional domain adapters (singing bowls, speaker enclosures)
• Web-based UI

---

📚 Citation

If you use this work in research, please cite:

@software{golden_studio_2025,
  author = {Cazzaniga, Alessio Ivoy},
  title = {Golden Studio: Evolutionary Optimization for Vibroacoustic Plates},
  year = {2025},
  url = {https://github.com/Alemusica/Rememberance}
}

---

📄 License

MIT License - see LICENSE

---

🙏 Acknowledgments

• pymoo team for NSGA-II implementation
• scikit-fem for finite element solvers
• All researchers whose papers informed this work (see bibliography)

---

"Geometry has two great treasures: one is the theorem of Pythagoras; the other, the division of a line into extreme and mean ratio (golden ratio). The first we may compare to a measure of gold; the second we may name a precious jewel." — Johannes Kepler

✦ φ ✦