Let's assume that we receive a and b, which are 1D signals of time T. At this time, if the relation of a(t) = k ⊗ b(t) holds, is there a way to find the deconvolution kernel k with a Neural Network (not an recursive method)?
This problem can be decomposed as B+ a = k after transforming b into a toplitz matrix and generating a pseudo-inverse B+ through decomposition methods such as SVD.
Is Neural Network possible to infer k after extracting the feature map between two vectors a, b and utilize the relationship (correlation or concatenation) between two vectors?
