MINE-Mutual-Information-Neural-Estimation-
MINE-Mutual-Information-Neural-Estimation- copied to clipboard
What is the optimal # of units in the intermediate layer for p and q variables?
I am implementing a mutual information estimate based on your coding. Suppose X and Y are in p and q dimensions, which means the input layer has p+q units. Then, how many units do you think are the most appropriate for the intermediate layer size? In fact, I took many days to search for the optimal solution. Do you have any idea?
Hi did you find any optimal solution for it? I have the same issue. specifically, with high dim of p,q
Hi! Do you have any solutions? I meet the same issue. Currently, I set it to 4*d.