Extended Data Fig. 1: Illustration of the iterative mechanism underpinning the DLVPM algorithm.
From: Integrating multimodal cancer data using deep latent variable path modelling
a: Shows the overall model. The aim of the method is to maximize the sum of correlations between deep latent variables (DLVs) connected by the path model. This optimization can be achieved by minimising a sum of least squares losses between the output of each measurement model, and the measurement models they are connected to via the path model. b: The overall loss can be minimized by iteratively minimizing least squared losses between the output of each measurement model, with outputs from connected measurement models.
