How to obtain the convex curve for weights vs loss in a neural network [closed]

Question

In most of the literature of Neural networks the 3D plot of weights, bias and the loss function is shown as below, When I tried I obtained a plot like this one

Here are more details, Here is the glimpse of the dataset, there are 15,000 training examples,

Here is the formula of the cost function used,

Here is the link to the ipython notebook with the code and visualization.

I want to obtain the curve similar to the one that's shown in the first image for demonstration purpose.

Answer 1

To obtain the convex curve, do the following:

Plot coefficient w (weights) in your x axis Plot intercept b in your y axis Plot the cross entropy in your z axis.

Define a range for w and b and calculate cross entropy for each couple w, b

Plot x,y,z values in 3D. It's done.