getLearningInsights

Provides comprehensive analytics and performance insights about the learning process.

This method aggregates training statistics, performance metrics, and behavioral analytics to provide detailed insights into the neural network's learning progress.

Analytics Components:

Core Metrics:

• interactionCount: Total number of processed user interactions

• progressEstimate: Learning progress as percentage 0.0, 1.0

• Based on interaction count with saturation at 100 interactions

Performance Analysis:

• averageConfidence: Mean confidence from recent training examples (last 20)

• trainingSteps: Total number of completed backpropagation iterations

• modelVersion: Architecture identifier for version tracking

Behavioral Insights:

• confidenceHistory: Complete prediction confidence timeline

• Enables trend analysis and performance visualization

• Sparkline data for monitoring prediction quality over time

Statistical Calculations:

averageConfidence = Σ(recent_confidence) / recent_count
progressEstimate = min(interactionCount / 100.0, 1.0)

Thread Safety:

• Uses model mutex for consistent snapshot capture

• Defensive copying of mutable collections

• Atomic read of all metrics

Performance Monitoring:

• Low overhead metric collection

• Efficient recent history windowing

• Memory-bounded confidence tracking

See also