Caffe for feed forward networks

Question

Has anyone experience with using caffe as feed forward network instead of convolutional neural network?

My input data is 1 dimensional. Everything is ok, but when I want to use layers like max pooling, caffe assumes a square shaped kernel size. I would rather need a 1 dim max pooling kernel.

Answer 1

Found it:

layer {
  name: "pool1"
  type: "Pooling"
  bottom: "ip1"
  top: "pool1"
  pooling_param {
    pool: MAX
    kernel_h: 3
    kernel_w: 1 
  }
}