Metrics Module

The Metrics module provides performance metrics for model evaluation.

For complete API reference, see the Doxygen-generated HTML documentation in doxygen-docs/html/.

Metric Implementations

Accuracy Metric

Classification accuracy metrics for model evaluation.

Location: src/AnalysisG/metrics/accuracy/

Classes: * accuracy_metric - Main accuracy metric implementation * collector - Data collection for accuracy analysis * cdata_t - Data structure for collected metrics * cmodel_t - Model-level metric aggregation

Features: * Edge-level accuracy * Node classification accuracy * K-fold validation support * Epoch-wise tracking * Mode-specific metrics

PageRank Metric

PageRank-based importance scoring for graph nodes.

Location: src/AnalysisG/metrics/pagerank/

Features: * Graph centrality measures * Node importance ranking * Iterative PageRank computation * Weighted graph support

Metric Interface

All metrics inherit from metric_template and implement:

define_metric() - Metric initialization
define_variables() - Variable setup
event() - Per-event computation
batch() - Per-batch aggregation
end() - Final computation

Usage Example

// Create and configure metric
auto* metric = new accuracy_metric();
metric->define_metric(&metric_data);
metric->define_variables();

// Process events
for (auto* event : events) {
    metric->event();
}

// Compute batch metrics
metric->batch();

// Finalize
metric->end();