Plotting loss/accuracy vs. epoch with PyKeras

Newbie
,
1

Dear experts,

I am working on building a neural net with the PyKeras method of TMVA. I would love to be able to make a plot of the accuracy and loss over the course of all the epochs something like this:

image

But I can’t figure out how, if it’s even possible. Is there a way to call the accuracy and loss values of each epoch so I can plot them?

ROOT Version: 6.18.00

2

I guess you looked at the TMVA tutorials and to the TMVA manual and did not find an answer.
@moneta can help I am sure.

3

Hi,

It should be visible in the TMVAGUI as item (8) , TrainingHistory.
You can start the TMVA GUI by doing from the ROOT prompt:

TMVA::TMVAGui("TMVA_Output.root");

where TMVA_Output.root is your output file of the training defined when you create the TMVA::Factory object.

Cheers

Lorenzo

4

Thanks Lorenzo,

Unfortunately my GUI looks like this, with no TrainingHistory button. Is there some option I need to have enabled, or maybe an updated version of ROOT?

image
image659×780 115 KB

5

Hi,
Which ROOT version are you using it ? Maybe it is too old and this was not yet implemented

Lorenzo

6

Yes , the TrainingHistory in the GUI has been introduced 2 years ago, from the ROOT version 6.20 and you are using 6.18, seeing from your first post.
If you can upgrade version, then you will have it

Cheers

Lorenzo