Suppose I have a
I want to construct a network (such as simple MLP) to fit
y1_train and to be low correlated with
y2_train (or to fit
How could I achieve this goal? Is the custom loss function a good solution?
I use keras as my tool.