Metrics Collection

The MetricsCollector uses Welford's algorithm for numerically stable running statistics.

Basic Usage

#![allow(unused)]
fn main() {
use entrenar::monitor::{MetricsCollector, Metric};

let mut collector = MetricsCollector::new();

// Record individual metrics
collector.record(Metric::Loss, 0.5);
collector.record(Metric::Accuracy, 0.85);

// Record batch of metrics
collector.record_batch(&[
    (Metric::Loss, 0.45),
    (Metric::Accuracy, 0.87),
    (Metric::GradientNorm, 1.2),
]);
}

Available Metrics

Getting Statistics

#![allow(unused)]
fn main() {
let summary = collector.summary();

if let Some(loss_stats) = summary.get(&Metric::Loss) {
    println!("Loss - mean: {:.4}, std: {:.4}", loss_stats.mean, loss_stats.std);
    println!("       min: {:.4}, max: {:.4}", loss_stats.min, loss_stats.max);

    if loss_stats.has_nan {
        println!("WARNING: NaN values detected!");
    }
}
}

Welford's Algorithm

The collector uses Welford's online algorithm for O(1) updates:

mean_new = mean_old + (x - mean_old) / n
M2_new = M2_old + (x - mean_old) * (x - mean_new)
variance = M2 / (n - 1)

This provides:

• Numerical stability for large datasets
• Constant memory usage
• O(1) per update