Geoffrey Hinton: Unpacking The Forward-Forward Algorithm
Eye On A.I.
The Correct Learning Signal for a Boltzmann Machine
When Boltzmann machines what you do is you give it positive data, real data, and you let it settle to equilibrium. And then in the negative fails, you do the same thing with stuff. You just let the model settle as producing data itself. That is the correct learning signal for a Boltzmann machine. The objective isn't to predict the next character. But having done that learning got the right representations for these strings of characters,. These windows of characters. In the negative phase they use characters that have been predicted already. So if there's a difference, then you'll be learning to make things more like the positive phase and less like the negative phase. If your predictions were
Remember Everything You Learn from Podcasts
