Extended Data Fig. 1: Diagram of the decoder architecture.
From: DynaMight: estimating molecular motions with improved reconstruction from cryo-EM images
The queried position is lifted to a higher dimensional space via a fixed positional encoding function, where the lifting dimension is defined by Np. The Nl dimensional latent code is concatenated with the encoded position, and input to a multilayer perceptron (MLP), which outputs a 3-dimensional displacement vector. To obtain the final position the displacement vector is added to the original position x.
