Best approach to using the fast.ai library for structured classification

Many thanks to Jeremy and Rachel for the fast.ai library, which makes using Pytorch easier - especially when using some of the learning rate tools built in.

I have been trying to use the ColumlarModelData method with a multi-class classification problem, as compared to the single output prediction of the Rossmann example.

What would be necessary to do this. It looks to me as though I would need to one hot encode the output for each class, is this the case? Also I was also wondering as to the best way to change the loss function.

Many thanks

John

3 Likes