4 ms·
This isn't true. A loss function that asymptotes can also have no minima, which commonly used loss functions do.
by moultano 6y ago
This isn't true. A loss function that asymptotes can also have no minima, which commonly used loss functions do.
- bananaface 6y agoDoesn't the fact the network is discrete (floats have a maximum precision) mean this isn't actually the case? There's a finite number of states the net can be in, and one (or more) is best.