p-koo / learning_sequence_motifs

"Representation Learning of Genomic Sequence Motifs with Convolutional Neural Networks" by Peter K. Koo and Sean R. Eddy
https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1007560
31 stars 13 forks source link

Input_shape #1

Closed MolGen closed 6 years ago

MolGen commented 6 years ago

Could you clarify what is the benefit of using input_shape = [None, 1000, 1, 4] instead of input_shape = [None, 1000, 4]?

p-koo commented 6 years ago

For historical reasons, I use conv2d. I will modify the code to use conv1d and a more concise shape in the future.