Question about logistic regression cost plot in C1_W3_Lab06_Gradient_Descent_Soln

[jiricojr_ttv]

One thing I couldn’t understand about the plot below: why does the chart show log(Cost(w,b)) instead of just Cost(w,b)? Doesn’t the logistic regression cost function already use logarithms?

[gent.spah]

The reason for using log of something in charts is to reduce the scale of the numbers in the graph so we can get visually a better understanding of what happens. This would normally be done when the values associated with each axis differ a lot in magnitude.

[conscell]

Without log scaling the contour plot would look like this:

[TMosh]

It is a convenience for visualization.