Validation loss higher than training loss from first epoch

Hi all,

From what I’ve learned when the validation loss > training loss there is overfitting. However I’m getting this from the first epoch. See below:

I’m trying to predict if a person is sitting, lying on the floor, standing, lying in bed or sitting in bed with a tabular learner that takes pose coordinates as inputs (x, y, confidence mixed with some room semantic).

I have 360K annotated poses, of which the majority is sitting, then standing, etc… We got about 20K lying on floor poses. I upsample all training data so all categories are equally represented. The validation set is 2% of the training set. I tried lowering the number of layers. This brings the training loss a bit closer to the validation loss, but the validation loss always is high than the training loss, from the first epoch on. Anyone have any explanation for this?

Hi rept hope all is well!

May be the following suggestions can be of help.

  1. Create a baseline model of classes with an equal number images in each class (no upsampling).
  2. Make the validation set 20% of the above data.
  3. Make a test set which is 10% of the above data.
  4. Make sure the classes have clean data in them e.g. no lying in the sitting class.
  5. Your baseline model could have 200-1000 items in each class as this is a test model.

If you train this model and it works fine, then this should help pinpoint any problems or errors that are occurring elsewhere.

Hope this helps.

Cheers mrfabulous1 :grinning: :grinning:

1 Like

Hi @rept and welcome!

It’s not correct that validation loss > training loss always means overfitting. You never want to have a model where training loss is > validation loss, that would mean you didn’t train enough and are underfitting. (See lesson 2 of last year’s course, starting from 49:00 to 53:00)

As long as your accuracy doesn’t get worse, you’re not overfitting :slight_smile:

2 Likes

Thanks for the answers!