I'm currently attempting to implement the a CNN for EEG classification using keras, specifically the one described in the paper here: https://arxiv.org/abs/2004.00077
It's all relatively straightforward until it states that they used a stride of 1 X 0 in the DepthConv2d layer. This throws an error with Keras, and according to all the sources I can find, none of the values in the stride parameter can be zero. When I ignore that, and set the stride to 1 X 1, it doesn't match the expected parameters from the document. Is there something I'm missing about this?
This is an excerpt of the paper showing what I'm trying to emulate. Fig. 3 and Table 1 from the paper
