After the learn model have been train and one wants to look at the results, is there a way for the method learn.show_results() to do it randomly?
2 Likes
I second this. With the full MNIST dataset, show_results displays only sevens. What would be useful is a representative sample of the results. With show_results, you are trying to assess whether the model is behaving sensibly. A totally biased sample is not very helpful.
How about a “shuffle” or “sample” boolean flag?
And, yes, I need to finally learn how to test and submit a PR.
I second that, it would be great to have a representation across the different classes in the dataset. I have not found yet in the framework how to do it