Does a CNN depend on the batch size?


I’m trying to grasp Conv1d. Now assume I have a data loader with a batch size N. If I create a sequential CNN, does the input layer and first CNN layer of that CNN depend on N?

AFAIK, conv layers don’t depend on N.
However, the batch norm would need a constraint that N != 1 during training.

Other than this, the training behavior might be different with different Ns, as the gradient will be more or less noisy depending on N.

Okay, thanks.

Edit: I’ll make a new thread with a follow up question. Thanks.