I am struggling with explaining the regular spikes in both training and validation losses observed when running the cats-and-dogs CNN from the book, see image below:
There is an interesting discussion of a very similar issue on SO:
, but my data size is an integer multiple of the batch size, so the explanation does not hold for me. I am unable to explain the sharp spikes and the steep drops in training loss at the end/beginning of each epoch.
I am struggling with explaining the regular spikes in both training and validation losses observed when running the cats-and-dogs CNN from the book, see image below:
There is an interesting discussion of a very similar issue on SO:
Why does my training loss have regular spikes?
, but my data size is an integer multiple of the batch size, so the explanation does not hold for me. I am unable to explain the sharp spikes and the steep drops in training loss at the end/beginning of each epoch.
I am happy to provide the notebook I used.
Thx!