How nathanwalash/comp3931_individualproject works
This repository is a transdermal diffusion research simulator and ML-surrogate workflow. The principal path runs configured simulations, validates progressively more complex regimes, assembles run outputs into train/validation/test data, and compares black-box and physics-corrected flux-curve surrogates. The graph includes solver internals, dataset handling, and model evaluation capabilities documented in the README. Some subsystem details are documented rather than sampled directly, so those links cite the README; unsampled implementation details are not inferred.
Main entry points: run_utils.py Run setup and outputs · train_corrective.py Corrective training · train_blackbox.py Black-box training · dataset.py Run generation and assembly
Components 24
Simulation
- Run configuration
src/skin_diffusion/config.py - Simulation runner
scripts/sim/run_sim.py - Run setup and outputs
src/skin_diffusion/run_utils.py - Diffusion solver
src/skin_diffusion/solver.py - Boundary conditions
src/skin_diffusion/bc.py - Diffusion operators
src/skin_diffusion/operators.py - Layered fields
src/skin_diffusion/layers.py - Transport metrics
src/skin_diffusion/metrics.py
Validation
- V1 numerical checks
scripts/sim/benchmark_v1.py - V2 literature calibration
scripts/sim/compare_literature.py - V3 regime validation
scripts/sim/validate_v3.py
Dataset pipeline
- Dataset specification
src/skin_diffusion/dataset_spec.py - Run generation and assembly
src/skin_diffusion/dataset.py - Run index
src/skin_diffusion/run_index.py - ML split preparation
src/skin_diffusion/ml_dataset.py - ML run loading
src/skin_diffusion/ml_run_dataset.py
Surrogate workflow
- Black-box surrogate
scripts/ml/blackbox_model.py - Corrective surrogate
scripts/ml/corrective_model.py - Black-box training
scripts/ml/train_blackbox.py - Corrective training
scripts/ml/train_corrective.py - Surrogate evaluation
src/skin_diffusion/ml_metrics.py - Physics and scalar diagnostics
src/skin_diffusion/ml_physics_diagnostics.py - Curve plots
src/skin_diffusion/ml_curve_plots.py
Other components
- Researcher
Connections 33
- Researcher → Simulation runner: starts run
README.md - Simulation runner → Run configuration: loads settings
scripts/sim/run_sim.py - Simulation runner → Run setup and outputs: runs simulation
scripts/sim/run_sim.py - Run setup and outputs → Layered fields: builds fields
src/skin_diffusion/run_utils.py - Run setup and outputs → Diffusion solver: calls solver
src/skin_diffusion/run_utils.py - Run setup and outputs → Boundary conditions: builds patch
src/skin_diffusion/run_utils.py - Diffusion solver → Boundary conditions: applies conditions
src/skin_diffusion/solver.py - Diffusion solver → Diffusion operators: steps diffusion
src/skin_diffusion/solver.py - Run setup and outputs → Transport metrics: derives transport
src/skin_diffusion/run_utils.py - Researcher → V1 numerical checks: checks numerics
README.md - Researcher → V2 literature calibration: checks literature fit
README.md - Researcher → V3 regime validation: validates final regime
README.md - V1 numerical checks → Diffusion solver: benchmarks
scripts/sim/benchmark_v1.py - V2 literature calibration → Transport metrics: compares targets
README.md - V3 regime validation → Run setup and outputs: validates regime
scripts/sim/validate_v3.py - Dataset specification → Run generation and assembly: samples parameters
scripts/sim/make_dataset.py - Researcher → Run generation and assembly: generates runs
README.md - Run generation and assembly → Run setup and outputs: generates simulations
src/skin_diffusion/dataset.py - Run generation and assembly → Run index: indexes runs
scripts/sim/check_runs.py - Run generation and assembly → ML split preparation: assembles splits
- ML split preparation → ML run loading: loads run bundles
src/skin_diffusion/ml_dataset.py - ML run loading → Run index: resolves run ids
src/skin_diffusion/ml_run_dataset.py - Researcher → Black-box training: trains baseline
README.md - Researcher → Corrective training: trains hybrid
README.md - Black-box training → Black-box surrogate: fits model
scripts/ml/train_blackbox.py - Corrective training → Corrective surrogate: fits model
scripts/ml/train_corrective.py - Corrective training → Black-box surrogate: uses backbone
scripts/ml/train_corrective.py - Black-box training → Surrogate evaluation: computes metrics
scripts/ml/train_blackbox.py - Corrective training → Surrogate evaluation: computes metrics
scripts/ml/train_corrective.py - Black-box training → Physics and scalar diagnostics: runs diagnostics
scripts/ml/train_blackbox.py - Corrective training → Physics and scalar diagnostics: runs diagnostics
scripts/ml/train_corrective.py - Black-box training → Curve plots: creates plots
scripts/ml/train_blackbox.py - Corrective training → Curve plots: creates plots
scripts/ml/train_corrective.py