On every iteration I get the same cost. The parameters are changing but almost appear to bounce around versus going to the boundary.
Now, if I assume my cost function is incorrect, that doesn’t appear to matter (as the update of parameters are not dependent on the cost function).
Check if your code is using the correct and consistent variable names for the cost.
I will check to ensure that my cost function is not using some variable from another cell and I’m using the function parameter.
But something that nags at me is that I should still be converging even if this function is incorrect … yet the parameters are not matching what they should be.
I had an issue in nn_model. I was calling backward_propagation incorrectly. So, in fact, the cost was not decreasing.
Good catch!