Explain error_rate as a metric

Is the default metric in the first few chapers which is error_rate based in percentages, such as an error_rate is 0.50 means 50% of your images correct?

Hey @ckehl
Yes, 50% error on the validation set!

error_rate = 1 - accuracy

accuracy: (pred == targ).float().mean()

context: binary

So, yes, it looks like it’s in percentage/ decimal form.

1 Like