Accuracy Formula in DLS?


#1

Hello,

How do we determine or find the accuracy formula that is used within DLS when applying any AutoML model? This is not observable in the Keras model code, so I am assuming it is taking place on the back-end with python during training. This is crucial for applied research, as the academics want to know this information. My thesis advisor is already asking me. :slight_smile:

Thanks,

Anthony


#2

Hi Anthony
For accuracy the formula used is
acc = Number of correct prediction/Total number of prediction
and no you can view the keras code of the model you have generated with autoML just click on the view option and select keras from the available option.

Regards
Rajat


#3

Thank you for confirm the formula Rajat; yes, I am aware I can via the Keras code via the method you have shown - what I am trying to understand is where is the “accuracy formula” taking place? I do not see it in the Keras code. Where is accuracy being computed?

Thanks,

Anthony


#4

Its computed after the model is created. The code which is generated by DLS is only for the model created step using DLS after which when you start training DLS automatically compile your model with other hyper partners and compute score and loss in the backend.