1-dimensional convolution: comparison with 'Matched filters'

Hi All!

I’ve recently discovered the fast.ai community and I am really impressed with the exceptional content, creativity and approach to learning.

I have a question which betrays my background:

  • if we use a 1-dimensional convolution neural network on a time series (which I am anticipating will mimic the classical linear time invariant models, since these are based on impulse responses and convolution).

Does target recognition with this kind of neural network reproduce the ‘Matched filter’? (Wiki’s examples are really nice)

Any direction in this line would be most useful.