Monte Carlo Analysis

The Monte Carlo analyzer runs repeated simulations with regenerated phase screens to compute statistical properties of the turbulent channel.

Usage

python gpu_main.py --mode monte_carlo

How It Works

The MonteCarloAnalyzer class (analysis/monte_carlo.py):

1. Runs num_runs independent simulations
2. Regenerates phase screens before each run (recreate_initial_ps())
3. Computes requested metrics per run
4. Saves checkpoints at configurable intervals
5. Aggregates statistics across all runs

Configuration

Set Monte Carlo parameters in config.json:

{
  "monte_carlo": {
    "enabled": { "value": true },
    "num_runs": { "value": 1000 },
    "save_interval": { "value": 100 },
    "analysis_metrics": {
      "value": ["scintillation_index", "mcf"]
    }
  }
}

Parameter	Default	Description
num_runs	1000	Total number of independent simulations
save_interval	100	Save checkpoint every N iterations
analysis_metrics	["scintillation_index", "mcf"]	Metrics to compute

Available Metrics

Scintillation Index

Measures intensity fluctuations:

SI = Var(I) / <I>^2

where I is the intensity (amplitude squared) and <I> is the mean intensity.

Mutual Coherence Function (MCF)

Computes the spatial coherence of the output field by averaging over multiple realizations using MCF_calc().

Output Structure

RUN_N/
  iterations/
    iter_0000/
      result.npy
    iter_0001/
      result.npy
    ...
  checkpoints/
    checkpoint_0100.json
    checkpoint_0200.json
    ...
  results/
    all_results.json
    statistics.json
    scintillation_histogram.png
    metric_evolution.png
    performance_metrics.png
    summary_report.txt

Statistics Output

The statistics.json file contains aggregated results:

{
  "scintillation_index": {
    "mean": 0.42,
    "std": 0.08,
    "min": 0.21,
    "max": 0.73,
    "median": 0.41
  }
}

Checkpointing

Intermediate results are saved every save_interval iterations. If a run is interrupted, previously completed iterations are preserved in the iterations/ directory.