Access and Sort Validation Set By Accuracy (Reverse of interp.top_losses)?

How can I sort the validation set by accuracy?

I am trying to plot the top and bottom classifications in the validation set.

We can access inter.plot_top_losses(n) but how about the opposite of it?


I think you’re asking two separate questions here. Your first one is

  1. Is it possible to use the Interpretation object to plot items by a metric, in this example accuracy, instead of by loss. For that one I don’t think there’s an easy answer other than looking at the code and then writing it yourself.

  2. You’re asking for the opposite of plot_top_losses(n), which would be something like plot the items with the lowest loss, instead of the highest loss. And for that, it’s easily done with:

interp.plot_top_losses(n, largest=False)
1 Like

This is exactly what I needed :slight_smile: Thanks a lot!