Myrl G. Marmarelis

I am serving as a postdoctoral scholar in the Computing + Mathematical Sciences department at the California Institute of Technology (Caltech). My mentors are Anima Anandkumar, R. Michael Alvarez, and Frederick Eberhardt. In May 2024, earned my PhD in Computer Science from the University of Southern California while working at the Information Sciences Institute under the supervision of Greg Ver Steeg, Aram Galstyan, and Fred Morstatter. My contributions continue to focus on robust causal inference and high-dimensional statistics. I have been fortunate to forge collaborations across multiple disciplines, including a project with Heinz-Josef Lenz using clinical-trial data, a project with Neda Jahanshad using the UK Biobank, and a project with Abigail Horn on mobility data.

I spent the summer of 2024 as a statistics engineer at Eppo, building a pipeline for long-term metric prediction in experiments. I had previous internships in machine learning and data engineering at Bloomberg and Research Affiliates.

Separately, I have spent time developing an end-to-end system for realtime monitoring of ultradian rhythms, those few-hour cycles in the human body that seem to modulate alertness. My eventual goal is to promote long-term wellness and health through novel uses of comfortable, noninvasive, and affordable wearables.

Selected Publications

M. G. Marmarelis, F. Morstatter, A. Galstyan, and G. Ver Steeg. Policy Learning for Localized Interventions from Observational Data, Artificial Intelligence and Statistics (AISTATS 2024 oral).

M. G. Marmarelis, R. Littman, F. Battaglin, D. Niedzwiecki, A. Venook, J.-L. Ambite, A. Galstyan, H.-J. Lenz, and G. Ver Steeg. q-Diffusion Leverages the Full Dimensionality of Gene Coexpression in Single-cell Transcriptomics, Communications Biology 7, 400 (2024).

M. G. Marmarelis, G. Ver Steeg, A. Galstyan, and F. Morstatter. Ensembled Prediction Intervals for Causal Outcomes Under Hidden Confounding, Causal Learning and Reasoning (CLeaR 2024 oral).

M. G. Marmarelis, E. Haddad, A. Jesson, N. Jahanshad, A. Galstyan and G. Ver Steeg. Partial Identification of Dose Responses with Hidden Confounders, Uncertainty in Artificial Intelligence (UAI 2023 oral).

M. G. Marmarelis, G. Ver Steeg, and A. Galstyan. A Metric Space for Point Process Excitations, Journal of Artificial Intelligence Research 73 (2022) 1323–1353.

M. G. Marmarelis and R. G. Ghanem. Data-driven Stochastic Optimization on Manifolds for Additive Manufacturing, Computational Materials Science 181 (2020) 109750.

Recorded Talks

Causality Discussion Group (August 2023) for the UAI '23 work.

UAI 2023 single-track oral presentation on Partial Identification of Dose Responses with Hidden Confounders.

USC Biostatistics Seminar (April 2023) on Latent Factor Discovery with Transcriptomics Data with Greg.

Aspirations

I am trying to make my work relevant in the battle against climate change, or improving public health. Feel free to reach out for possible collaborations.

Open Source

RankedChoices.jl --- A Julia package to facilitate analysis of ranked preferences.

rolling-quantiles --- A Python package to quickly stream rolling quantiles via a backend written in C.