Title: From toy models to Cryo-BIFE
Abstract: Toy models are extremely valuable, as they let us understand what is going on underneath the methods we develop. In this talk, I will go through a series of toy models we applied to our method for biasing molecular dynamics simulations with 2D images. With these toy models, we formulated the correct posterior for setting up the bias in the case of a heterogeneous dataset, which showed us how close our method was to Julian's Cryo-BIFE. Finally, I will show how we will use molecular dynamics to optimize the path found by Cryo-BIFE and how to think about the Cryo-EM imaging process algorithmically.