gnn-statistical-analysis

To aggregate GNN execution statistics across multiple frameworks, apply this analysis to simulation output directories. It extracts and aggregates metrics like entropy, mutual information, and complexity to enable cross-model comparisons, returning structured JSON outputs and summary metrics.

For GNN models from simulation results, you can compute cyclomatic, cognitive, and structural complexity metrics. The analysis also calculates quality indicators such as maintainability and technical debt to produce comprehensive quality assessment reports.

Yes, you can generate publication-ready visualizations and dashboards from GNN analysis outputs. The pipeline aggregates variable and connection statistics, returning structured JSON and summary metrics specifically suitable for visualization dashboards and research reports.

Yes, you need pandas, scipy, numpy, and matplotlib installed to run cross-framework GNN statistical comparisons. These dependencies support data aggregation, statistical computation, and visualization for frameworks like PyMDP, JAX, RxInfer, and PyTorch.

The best way to quantify entropy and mutual information for GNN execution traces is to run this statistical analysis on simulation results. It identifies key statistical properties and aggregates these metrics for cross-model quality assessment.