Affiliates

Faculty/Senior Researchers

Ryan Bennett (Linguistics)

Prof. Bennett is a faculty member in UCSC Linguistics. His research deals with quantitative and qualitative aspects of sound patterns in the world’s languages, with a particular focus on underdocumented languages of Latin America, as well as Celtic languages. Prof. Bennett’s work in this area draws on experimentation, corpus analysis, and traditional pen-and-paper methods alike.

Adrian Brasoveanu (Linguistics)

Professor Brasoveanu is a formal semanticist and computational psycholinguist. His research focuses on two questions: (i) What is linguistic meaning? (ii) How does the human mind grasp it? He pursues answers to these questions by building evidence-based, formally / computationally explicit theories of linguistic meaning (product) and interpretation (process). The empirical evidence comes from a variety of experimental tasks (forced choice, acceptability, self-paced reading, eye-tracking) and natural language corpora. The theory-building tools are both symbolic / ‘qualitative’ — mathematical logic, formal grammars — and subsymbolic / ‘quantitative’ — Bayesian and, less frequently, frequentist models for continuous and categorical data, e.g., mixed effects generalized linear models, mixture models, non-parametric models, neural-network models. For more information, see https://people.ucsc.edu/~abrsvn/.

Kevin Bundy (Astronomy & Astrophysics)

Kevin Bundy’s group is exploring machine learning applications to integral-field spectroscopic observations of galaxies. A key goal is to marshall high-dimensional data sets like MaNGA to gain physical insight into complex systems, like ongoing galaxy mergers, that are difficult to model with traditional techniques. He is also helping lead development of next-generation, high-multiplex spectroscopic facilities for Keck Observatory. Such facilities will provide the training sets necessary for data science applications in the coming era of “Big Data” astronomy, with particular promise for extracting physical information from panoramic deep imaging surveys like LSST.

Mentees: Grecco Oyarzun (PhD), Namrata Roy (PhD), Brian DiGiorgio (PhD), Viraj Pandya (PhD), Marina Huang (undergrad)

Angus Forbes (Computational Media)

Professor Forbes is exploring opportunities at the intersections of machine learning, graphics, virtual reality, and visualization. See https://creativecoding.soe.ucsc.edu for more information. 

Example publications:
  Deep Illumination: Approximating Dynamic Global Illumination with Generative Adversarial Networks
Text Annotation Graphs: Annotating Complex Natural Language Phenomena
NeuroCave: A Web-based Immersive Visualization Platform for Exploring Connectome Datasets
  Visualizing and Verifying Directed Social Queries
Transmitting Narrative: An Interactive Shift-summarization Tool for Improving Nurse Communication

Joshua Deutsch (Physics)

Professor Deutsch chas broad interests in theoretical condensed matter, biophysics, and statistical mechanics. He has used machine learning in a number of problems, such as the diagnosis of cancer by analysis of microarrays, music recognition systems, financial market prediction, and is currently interested in the relationship between the computational machinery of the genome and deep learning algorithms.

Sample publications:
Mentees engaged: Thomas L Madden,  Stephen Martin

Daniel Friedman (Economics)

Professor Friedman’s current research projects include (a) high frequency trading (HFT) in financial markets; many HFT algorithms use various forms machine learning. This work is joint with my UCSC Economics colleagues Eric Aldrich and Kristian Lopez-Vargas; and  (b) economics of e-commerce and recommender systems, including finding best-fitting parameter values for product choice via machine learning. This work is joint with BSOE colleague Yi Zhang.

See https://leeps.ucsc.edu/papers/ for samples.

Marcella Gomez (Applied Math)

Professor Gomez uses neural networks to model cellular response to diverse environmental conditions in order to elicit desired behavior. Single layer networks are used to generate predictive models and controllers with no memory. We are now looking into combining this technique with deep learning in order to add “intuition” into our model and controller.

Mentees engaged: Jianhong Chen (PhD student), Mohammad Jafari (Postdoc)

Alexie Leauthaud (Astronomy & Astrophysics)

Professor Alexie Leauthaud uses a range of machines learning techniques to analyze state of the art cosmology surveys such as HSC and DESI. She is also involved in future cosmology programs such as LSST and WFIRST.

Mentees: Song Huang, Enia Xhakaj, Chris Bradshaw, Felipe Ardila, Yifei Luo

Joel Primack (Physics)

Joel Primack’s group is using machine learning for understanding galaxy formation and evolution and also for improving our understanding of the evolving large-scale structure of the universe. One recent major project (Huertas-Company, Primack, et al. 2018, Astrophysical Journal 858, 114) taught a CNN to identify three stages of galaxy evolution using realistic simulated images of galaxies each tagged with one of three stages, and then used the trained CNN to recognize three stages for the entire dataset of Hubble Space Telescope images of distant galaxies. Our conclusion was the the HST images showed the same three stages, with the transitions occurring at the same galaxy stellar masses as our simulations. Recent papers, led by UCSC grad student David Reiman, use a GAN to de-blend overlapping galaxy images, and a CNN to improve the resolution of galaxy images. Other work in progress uses a CNN to discriminate between images of interacting galaxies and chance superpositions of galaxy images. Another project uses a GAN to make optimal use of a combination of mostly inaccurate photometric redshifts plus about 10% accurate spectroscopic redshifts to determine the environments of distant galaxies. Our group’s work has been supported in part by a 2016 grant from Google and a 2018 Google Faculty Research Grant to Primack, and Google machine learning experts have been very helpful to us.

J. Xavier Prochaska (Astronomy & Astrophysics)

Professor Prochaska applies a range of machine learning and deep learning (e.g. CNN, GAN) techniques to large datasets of astronomical spectroscopy.

Example publications:
  Deep learning of quasar spectra to discover and characterize damped Lyα systems
  Spectral Image Classification with Deep Learning
Mentees: Joe Burchett (postdoc), Zheng Cai (postdoc), Jiani Ding (PhD)

Stefano Profumo (Physics)

Professor Profumo’s group uses convolutional neural networks (CNNs) to understand the spectra and morphology of gamma-ray emission, in the search for possible signals of new physics such as from particle dark matter; group affiliates are also working on a CNN that will help users discriminate between point-like and extended gamma-ray emission, and on “bump hunting” in both gamma-ray spectra and in high-energy physics data sets such as those generated by the Large Hadron Collider.

Mentees: Grad students Jaryd Ulbricht and John Tamanas; undergrad students John McMurry and Jeremy Diamzon-Larot

Kyle Robertson (Philosophy)

Kyle Robertson is the Assistant Director of the Center for Public Philosophy and a lecturer in the philosophy department. His research and teaching deal with ethics and applied ethics, and much of his work is in public philosophy, where he engages in philosophical dialogue with communities around the country. He is interested in the ethical issues that arise in AI research and technology, and the legal and political ramifications of AI (before coming to the study of philosophy, Kyle was an IP litigator at a large Silicon Valley law firm).

Amanda Rysling (Linguistics)

Professor Rysling is an assistant professor in the Department of Linguistics at the University of California, Santa Cruz. She is interested in how language is represented and processed; in particular, cases in which the processing of a linguistic unit is affected by the phonetic, phonological, or prosodic context in which that unit occurs. This theme unifies my work on segmental perception, lexical bias effects, cues to prosodic structure, and sonority sequencing and vowel alternations in Polish.

Matt Wagers (Linguistics)

Professor Wagers is a linguist with a focus on language comprehension and on how humans encode and remember syntactic information. Topics include the interpretation of non-local dependencies and the use of morphological information to manage short-term memory. An important theme is incorporating language processing data from underrepresented languages, including Chamorro and Zapotec. More information: https://people.ucsc.edu/~mwagers/.

 

Skip to toolbar