This is a repository for the codes used in the paper: DGP-LVM (https://arxiv.org/abs/2404.04074). The codes are arranged as follows.
DGP-LVM model development files for derivative squared exponential (SE), Matern 3/2 (M32) and Matern 5/2 (M52) covariance functions. Model development is perfomed using Stan (https://mc-stan.org/).
Simulation study designed to validate and compare performance of DGP-LVM and other GP models in estimating latent input variables. Two simulation scenarios have been considered: data generated from a GP and data generated from a periodic function.
This is to check model convergence using MCMC convergence diagnostics. The plots have been generated using the associated code.
Code for generating the plots showing recovery of ground truth latent inputs along with model hyperparameter recovery.
DGP-LVM is used to analyze a reduced single-cell RNA sequencing data. The data is a pre-processed Cell Cycle data obtained from Cytopath (https://doi.org/10.1016/j.crmeth.2022.100359).