Can someone explain how 'One-Dimensional Convolutional Neural Network' works. I do understand the 2-D for image but for 1-D how is the filer created. is it fixed 1-D filter within a specific time interval or the operation is the same as we convolve a signal with a filter in signal processing y = f*x


is it fixed 1-D filter within a specific time interval?

Yes. The same as filters in 2D. Adjacent filters may even have no overlap with each other.

1D CNN is almost the same as 2D CNN both mathematically and visually by setting the second dimension (either the horizontal or vertical one in visualizations) to 1. This way, 1D filters are placed (possibly with some overlap) in one dimension instead of 2D filters being spread in two dimensions.

The below image shows a filter set with shared parameter $W$ covering the overlapping regions of the input.

By shared parameter we mean $f_i=\mbox{ReLU}(\mbox{sum}(W \odot \mbox{region}_i))$, where $\odot$ is a point-wise product between a region of input and $W$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.