4 ms·

I assume that they're assuming a particular distribution of the target function when making this claim. Otherwise you have No Free Lunch. I wonder what that dis

• andy_ppp 3y ago
  Think of all solutions to a problem as a weirdly shaped wavy graph, you might select a solution that has some states that appear to work great but cause other better states to be impossible, the gradient search is good but you that and big jumps occasionally to check you aren’t in a local peak or trough. There’s loads of stuff in operations research about it.