Machine Learning

Zwicky Transient Facility (ZTF)

The Zwicky Transient Facility (ZTF) is a 47 square degree field of view time-domain survey from the Samuel Oschin 48-inch Schmidt telescope at Palomar Observatory in California. DR2 was released recently (Dec 2019) and boasts of 100 billion source detections in g/r/i filters combined. I lead the machine learning (ML) for the survey, and the team works on various classification aspects e.g. separating real-bogus sources, identifying streaking asteroids, pigeon holing sources in to their subclasses using advanced statistical and computational methods.

Catalina Real-time Transient Survey (CRTS)

The Catalina Sky Survey (CSS) NEO project uses three dedicated telescopes to cover thirty three thousand square degrees (now two in Arizona). The Catalina Real-Time Transient Survey (CRTS) utilized the CSS data to search for rare and interesting transients and variables. Currently the transients are not actively screened. I worked on classification of transients as well as building some initial infrastructure. keyword crts keyword catalina --

Early Detection Research Network (EDRN)

The Early Detection Research Network (EDRN), an initiative of the National Cancer Institute (NCI), working with tens of institutions on accelerating the translation biomarker research for clinical applications and to detect cancers as early as possible. I work with the Data Science team at JPL to do machine learning for early detection of cancer, and to help build datasets for that purpose. I also do methodology transfer applying techniques from astronomy to medical data to understand the patterns therein.

Molecular and Cellular Lesions (MCL)

Molecular and Cellular Lesions (MCL) or actually the mouthful Consortium for Molecular and Cellular Characterization of Screen-Detected Lesions (MCL) is made of teams studying tumor tissue, cell, and microenvironment components to distinguish screen-detected early lesions from interval and symptom-detected cancers. There is a lot of overlap of my EDRN work, and involves doing machine learning and dataset building for somewhat related aspects.

Automatic Learning for the Rapid Classification of Events (ALeRCE)

Automatic Learning for the Rapid Classification of Events (ALeRCE) is an effort for the classification of astronomical alerts. A majority of the members are from Chile, and as the ALeRCE team gets ready for LSST, they are using the ZTF alert stream for training models and algorithms. As part of CRTS and ZTF I have contributed some labeled data, and am loosely involved in the machine learning and classification effort of the group through collaborative meetings and through teaching at the La Serena Schools on Data Science (partly funded by NSF).