3 ms·

Or just use non-gradient based optimisation? Maybe I'm missing something.

• geysersam 4y ago
  That's usually really inefficient. But I agree that's the right direction (pun intended). Question is: How can we extract similar amount of information about the problem without relying on derivatives?