How biswajyoti-nath/constraint-interactions-symbolic-regression works
The product is a research framework that compares a theoretical constraint-interaction metric with empirical symbolic-regression speedups. The principal workflow runs metric, experiment, then analysis phases. The graph reflects the README and sampled modules; exact internal wiring for unsampled diagnostics and some experiment details is therefore shown only where documented.
Main entry points: main.py Phase orchestrator · analysis.py Statistical analysis · experiments.py Benchmark runner · metric.py Density and interaction
Components 14
Pipeline control
- Phase orchestrator
main.py
Metric estimation
- Constraint checks
src/constraints.py - Expression generator
src/expr_generator.py - Density and interaction
src/metric.py
SR experiments
- Benchmark runner
src/experiments.py
Analysis and diagnostics
- Statistical analysis
src/analysis.py - Monte Carlo ablation
scripts/mc_ablation.py - Failure predictor
scripts/failure_mode_predictor.py - Distribution audit
scripts/distribution_audit.py
Other components
- Researcher
- Configuration
config.yaml - PySR engine
- Benchmark datasets
- Analysis outputs
Connections 17
- Researcher → Phase orchestrator: selects phase
- Phase orchestrator → Configuration: reads settings
- Phase orchestrator → Constraint checks: builds checks
- Phase orchestrator → Expression generator: generates expressions
- Phase orchestrator → Density and interaction: computes metric
- Phase orchestrator → Benchmark runner: runs benchmarks
- Phase orchestrator → Statistical analysis: runs analysis
- Density and interaction → Constraint checks: evaluates expressions
- Density and interaction → Statistical analysis: provides interaction matrix
- Benchmark runner → Constraint checks: checks compliance
- Benchmark runner → Benchmark datasets: loads benchmark data
- Benchmark runner → PySR engine: runs regression
- Benchmark runner → Statistical analysis: provides experiment results
- Statistical analysis → Analysis outputs: writes summaries and figures
- Monte Carlo ablation → Density and interaction: sweeps sample size
- Failure predictor → Statistical analysis: predicts failure modes
- Distribution audit → Statistical analysis: audits distributions