Hi
I am currently trying to interpret the predictions of a fully connected Neural Network on a regression task. The inputs are adjacency matrices for a brain neural networks (from diffusion MRI scans) to predict cognitive abilities. My model currently does a good job at predicting (mean error of 5%) - I would like to understand why it is making the decisions it is making (e.g. which regions of the brain are more important).
I have come across a lot of literature covering this problem, but only regarding constitutional neural networks, not fully connected ones.
Thanks!