Aditya’s interests are mainly in probabilistic machine learning and he works on probabilistic interpretations of dimensionality reduction methods used in single cell biology. His current work interprets a large class of DR methods such as (t-)SNE and UMAP, LLE, LE, diffusion maps and spectral embedding methods as inference algorithms corresponding to models on latent graphs, with connections to Gaussian processes on graphs and manifolds. He also works on several case studies of using latent variable models in practice.
Gaussian Process Latent Variable Flows for Massively Missing Data
Third Symposium on Advances in Approximate Bayesian Inference, :