But what does Knowing Why The Network Works mean exactly? I would say it means: Why do these algorithms seem to be doing "well" despite their well-known theore

But what does Knowing Why The Network Works mean exactly? I would say it means: Why do these algorithms seem to be doing "well" despite their well-known theoretical intractability? Is it something about the instances? Or is it simply a matter of scale (so the local optima they are getting stuck in are not as obvious)? Or is there something "deep" we do not understand yet about these algorithms (or the theory)? Either way, there is something going on here for sure.