Performance issues - Githubissues

davhbrown / interactive_classification_metrics

Get an intuitive sense for the ROC curve and other binary classification metrics with interactive visualization.

MIT License

6 stars 0 forks source link

Open davhbrown opened 10 months ago

davhbrown commented 10 months ago

Performance/responsiveness of interactive graphs gets worse as the number of samples (N) for either of the normal distributions increases.

this depends partially on device capabilities
this may improve by initializing each figure with output_backend="webgl", but requires an adequate GPU
bokeh 3.0 tests resulted in worse performance than bokeh 2.4.2

Current workarounds: keeping N low or clicking slider locations rather than sliding.

davhbrown commented 4 months ago

Update improved workaround: cap the max number of points generated by capping both "N" Sliders at 500 as of PR #6