Studying neural network loss landscapes provides insights into the nature of
the underlying optimization problems. Unfortunately, loss landscapes are
notoriously difficult to visualize in a human-comprehensible fashion. One
common way to address this problem is to plot linear slices of