AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is This a Good Way to Think About G Flow Net?
The idea was that if the pegs on the galton bord are precisely and symmetrically arranged, the beads will form a nice binomial curve at the bottom. And it seems like what g flow nets are capable of doing when they optimize the pathwaights. I they're tweaking the pegs a little bit to the left or a little to the right to bias the flow of beads one way or the other. In this way, a g flo net could arrange the pegs so that the beads could form any distribution at the bottom that we want. For our purpose, that means a distribution that matches the reward function.