How much overfitt do you tolerate?


Basically the title sums it up… Usually we battle ovefitt, especially with DL methods… Lets say one puts away 2020 data of certain stock. (lets pretend there was no corona outlier for the sake of the question).

Now imagine you have three models, with the following metrics:

  • 80% train accuracy, 60% test accuracy
  • 72% train, 67% test
  • 65 %train, 64 %test

Now one build a strategy on top of the model, so train overfitt could be an issue..

Which model would you prefer in production and why? (second or third, i put the first one just for perspective)

Submitted October 22, 2020 at 11:05AM by mlord99

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s