Model Validation

Examples of validation tests in modeva.

Model Residual

Residual analysis for classification and regression.

Residual Analysis (Classification)

Residual Analysis (Classification)

Residual Analysis (Regression)

Residual Analysis (Regression)

Model Performance

Performance metrics and sliced performance analysis.

Performance Metrics (Regression)

Performance Metrics (Regression)

Performance Metrics (Classification)

Performance Metrics (Classification)

Sliced Performance (Regression)

Sliced Performance (Regression)

Sliced Performance (Classification)

Sliced Performance (Classification)

Overfit Detection

Detecting overfitting via sliced diagnostics.

Overfitting Analysis (Regression)

Overfitting Analysis (Regression)

Overfitting Analysis (Classification)

Overfitting Analysis (Classification)

Reliability Analysis

Prediction reliability and uncertainty analysis.

Reliability Analysis (Regression)

Reliability Analysis (Regression)

Reliability Analysis (Classification)

Reliability Analysis (Classification)

Resilience Analysis

Model resilience under distribution shift.

Resilience Analysis (Classification)

Resilience Analysis (Classification)

Resilience Analysis (Regression)

Resilience Analysis (Regression)

Robustness Analysis

Model robustness to input perturbations.

Robustness Analysis (Regression)

Robustness Analysis (Regression)

Robustness Analysis (Classification)

Robustness Analysis (Classification)

Fairness Analysis

Model fairness analysis.

Model Fairness Analysis (Classification)

Model Fairness Analysis (Classification)

Explainability

Global and local post-hoc explainability.

Local Explainability

Local Explainability

Global Explainability

Global Explainability

Weakness Region Detection

Gallery examples for AMIF weakness region diagnostics, which identify regions where a model performs poorly using geometry and mutual information scoring.

Weakness Region Analysis (Regression)

Weakness Region Analysis (Regression)

Weakness Region Analysis (Classification)

Weakness Region Analysis (Classification)

Weak-Cluster Detection and Repair (FuseKernel + MoE)

Weak-Cluster Detection and Repair (FuseKernel + MoE)