Stephen Portillo

Stephen is interested in developing data analysis techniques using the latest advances in statistical inference in order to leverage complex astronomical data. In his thesis, Stephen developed probabilistic cataloging, a novel Bayesian technique that is able to handle extremely crowded images of stellar fields. The method is able to handle deblending ambiguities by treating the number of sources itself as a parameter to be inferred using trans-dimensional Markov chain Monte Carlo sampling. He is also interested in using variational autoencoders as a dimensionality reduction technique for high-dimension observations like optical galaxy spectra.