AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Limits of Normalized Important Samples
This idea of the self-normalized important sampling was cool as well. It seems like you were using it at multiple levels. And then I think in order to estimate the proposal distribution, you were then using self- normalized important sampling again. Yes. So it's like, not only are you improving the learning algorithm, you're also providing unification of the ideas.