Ludwig Experiment
Visualizations
confusion matrix AGE CAT top2
confusion matrix entropy AGE CAT top2
frequency vs f1 AGE CAT
learning curves AGE CAT accuracy
learning curves AGE CAT loss
roc curves
Feature Importance
Feature importance scores come from Ludwig's Integrated Gradients explainer. It interpolates between each example and a neutral baseline sample, summing the change in the model output along that path. Higher |importance| values indicate stronger influence. Plots share a common x-axis to make magnitudes comparable across labels, and the table columns can be sorted for quick scans.
| label | feature | importance | abs importance |
|---|---|---|---|
| post | RP11-465B22_8 | -0.007552 | 0.007552 |
| post | RP11-206L10_9 | -0.004579 | 0.004579 |
| post | RPL23AP24 | 0.003386 | 0.003386 |
| pre | RP11-465B22_8 | 0.007552 | 0.007552 |
| pre | RP11-206L10_9 | 0.004579 | 0.004579 |
| pre | RPL23AP24 | -0.003386 | 0.003386 |
Top features for post
Top features for pre