Issue with scale on validation error and accuracy graphs


While training the training and validation error are displayed on the same graph with iteration as the X axis. Same for training and validation accuracy:

In the result tab the validation error are displayed on a different scale, which looks like nepoch for the X axis:

It would be clearer if a X-axis was displayed underneath each graph.


Yes. We have received this feedback and plan to show the axis information in the graphs.


That should make the display more readable.

One more little thing is that there are 3 lines describe in the caption:
but only two are actually displayed:


‘Accuracy’ is the accuracy for each batch while AvgAccuracy is average of batch accuracy within one epoch.

‘Accuracy’ is hidden by default but will be shown when you click on it. It can be useful to know how model is doing on every batch (technically UI won’t be showing each batch as UI updates is limited to 1 per second).

If you see a very high and low in batch accuracy, the should give the indication that on some batches model is performing poorly.


I don’t quite understand what is plotted on the X-axis. Is it the time or the number of data passed. I also can’t deduce it from the past time and my number of data because it doesn’t fit.

Thank you very much,