How p2ms-optimization/pysne works
PySNE is a numerical-search library whose main workflow runs bounded candidate-region discovery, refines each region with Spiral Optimization, and filters the resulting candidates into roots or optima. The graph also captures its problem families, benchmark collections, and plotting capability. The solver and algorithm links are supported by sampled source; unsampled benchmark and plotting internals are represented only at paths confirmed by the tree and README.
Main entry points: base.py Problem models · solver.py Solver entry · modified_clustering_process.py Region clustering · engine.py SPO refinement
Components 16
Search workflow
- Solver entry
pysne/solver.py
Search algorithms
- Region clustering
pysne/clustering/modified_clustering_process.py - Cluster model
pysne/clustering/model.py - Sobol sampling
pysne/initialization/sampling.py - SPO refinement
pysne/optimizers/spo/engine.py - Rotation matrix
pysne/optimizers/spo/matrix.py
Problem definitions
- Problem models
pysne/problems/base.py - SNE benchmarks
pysne/problems/benchmarks_sne.py - Multimodal benchmarks
pysne/problems/benchmarks_multimodal.py - Diophantine benchmarks
pysne/problems/benchmarks_diophantine.py - Fitness and filtering
pysne/utils.py
Results and plots
- SNE plots
pysne/visualization/sne.py - Multimodal plots
pysne/visualization/multimodal.py - Roots or optima
Other components
- Library user
- Matplotlib
Connections 21
- Library user → Solver entry: calls
- Solver entry → Region clustering: discovers regions
- Region clustering → Sobol sampling: initializes points
- Region clustering → Rotation matrix: builds dynamics
- Region clustering → Problem models: evaluates fitness
- Region clustering → Cluster model: creates clusters
- Solver entry → SPO refinement: refines clusters
- Solver entry → Sobol sampling: samples refinements
- SPO refinement → Rotation matrix: builds dynamics
- SPO refinement → Problem models: evaluates objective
- Sobol sampling → Problem models: uses domain bounds
- Problem models → Fitness and filtering: computes and filters
- SNE benchmarks → Problem models: defines problems
- Multimodal benchmarks → Problem models: defines problems
- Diophantine benchmarks → Problem models: defines problems
- Solver entry → Problem models: selects candidates
- Solver entry → Roots or optima: returns
- Roots or optima → SNE plots: plots roots
- Roots or optima → Multimodal plots: plots optima
- SNE plots → Matplotlib: renders with
- Multimodal plots → Matplotlib: renders with