I will be working with the University of Michigan M-Write initiative that incorporates writing-to-learn pedagogies into university wide gateway STEM courses. Writing-to-learn engages students by asking them to write about science topics, to interact with one another through peer review, and to learn through a revision process. While research widely supports the benefits of writing and peer review in the learning process, writing in large gateway courses amasses a corpus of essays for which educators cannot quickly provide feedback. The M-Write initiative strives to build a digital toolkit supporting the application of writing-to-learn pedagogies at scale. I am working with M-Write to combine conceptual writing prompts, automated peer review, natural language processing, and automated personalized feedback to create an infrastructure for writing at scale. This will expand the use of content-focused writing activities to engage students in course material and thereby increase their comprehension of key concepts.
The field of computational fluid dynamics (and more broadly, computer simulations) gives us tools to help understand and study scientific/engineering problems without the need to run many costly, real life experiments. However, because the mathematical models underlying these simulations are (often) simplifications of the real processes of interest, their modelling/predictive power depends on the simulation’s resolution, and time or spatial scales of the problem of interest. During my MIDAS Postdoc, I will work towards developing techniques for multi-scale modelling through assimilation of experimental data (to account for the simplifications of the models), and towards developing machine learning techniques compliant with a set of constraints (for example, physics laws).
My primary focus is on developing foundational theory and methodology for data science using information theory, machine learning and signal processing. More specifically, my data science research aims to develop theory and algorithms for analysis of medical Big Data using Data In Motion with applications in Digital Health. In particular, my planned research objective is to design robust and validated clinical, physiologic, cellular, and genomic predictive of infection, as defined by viral shedding. In particular, we will identify parameters that predict a person’s contagiousness at the earliest possible time following exposure using genomic data and real-time physiological signals recorded by wearables (Empatica, Fitbit, UnitedHealth). As such, we design novel approaches tailored to complex data (high-dimensional, missing values, time-series, multi-modality) and identify low dimensional biological signatures characterizing the host response of individuals to virus inoculation. At the completion of this study we will have developed a model for contagion that will have significant military and public health impact, since soldiers as well as the general public pose infectious risks to those around them. The data generated will have collateral benefit for the scientific community investigating host-pathogen interactions and for the diagnostics and pharmaceutical communities for development of platforms to diagnose and treat pre-symptomatic infectious disease.
Patrick’s research interest centers on the structure and evolution of large-scale human social networks. His recent work aims to address how social networks adapt in the long term to heightened uncertainty caused by sudden and often unforeseen societal shocks, such as economic busts, corporate M&As and scandals, geopolitical conflicts, and natural disasters. When an area experiences a devastating storm, for example, do people turn to friends and family or acquaintances for information and support? How does the structure of people’s networks influence their communication patterns in response to such societal shocks? How do social networks respond differently depending on the nature of the shock (e.g., political coups vs. hurricanes)? These questions hold increasing significance as societies face mounting uncertainties due to climate change and computation-driven transformations in labor markets and industries. Understanding how social networks change amid heightened uncertainty, then is key to understanding and predicting important social, political, and economic processes that are shaped by social networks, from information diffusion, political polarization, to technological innovation. Patrick investigates these questions around uncertainty and network change, first, by empirically exploring the change and recovery of interpersonal communication networks after societal shocks using social media data (e.g., Twitter) and, second, by devising computational models to theorize the macro-structural implications of the changes in individual-level communication behaviors induced by the shocks.
With the explosion of the volume and variety of epidemic related data, we propose an efficient, theoretically sound, and easily-implemented framework, which employs both the regularly reported infections and the asynchronous genetic samples to make inference on the epidemic parameters (eg. transmission rate) and forecast the epidemic trends and genealogical patterns.
S Zhao, et al. Preliminary estimation of the basic reproduction number of novel coronavirus (2019-nCoV) in China, from 2019 to 2020: A data-driven analysis in the early phase of the outbreak. Int J Infect Dis. 2020 Mar;92:214-217. DOI: 10.1016/j.ijid.2020.01.050. Epub 2020 Jan 30.
S Zhao, et al. Estimating the Unreported Number of Novel Coronavirus (2019-nCoV) Cases in China in the First Half of January 2020: A Data-Driven Modelling Analysis of the Early Outbreak. J Clin Med. 2020 Feb 1;9(2). pii: E388. DOI: 10.3390/jcm9020388.
Q Lin, et al. A conceptual model for the outbreak of Coronavirus disease 2019 (COVID-19) in Wuhan, China with individual reaction and governmental action. Int J Infect Dis. 2020. DOI: 10.1016/j.ijid.2020.02.058.
My research is focused on cosmology and its intersection with fundamental physics. To study the constituents of our universe and demystify the small-scale astrophysics, I develop novel inference models and computational algorithms. As an astronomer, I employ the most massive gravitationally bound objects in the Universe, i.e. galaxy clusters, to study the nature of dark matter and dark energy. I am also an advocate for applications of data science for problems with societal impact. When I am not doing astronomy, I engage with policy- and decision-makers and enable them with data-informed decision making by providing novel data-driven tools.
I am an active member of several international projects and collaborations, including the Dark Energy Survey(DES), the COsmostatistics INitiative (COIN), XMM-XXL Consortium, among others. I was a McWilliams Postdoctoral Fellow at CMU, recipient of the best student paper award in KDD’18, an awardee of the Michigan Institute for Computational Discovery and Engineering (MICDE) fellowship, and recipient >$50k grants.
Undergraduate/graduate students: I am continually looking for dedicated undergraduate and graduate students (you) who want to expand their portfolio and take part in data science with social impact or astronomy projects. These projects involve a balance of theoretical, methodological, and data analysis work. If you are looking for a project feel free to email me.
World Economic Forum Report: I contributed to a report from the World Economic Forum featuring a data science project co-funded by MIDAS, U-M Transportation Research Institute (UMTRI), U-M College of Engineering and The Knight Foundation. The project is part of a larger Seamless Integrated Mobility effort that aims to transform mobility systems in Detroit, Ann Arbor and Windsor. The project is one example of how data science can make a significant impact on policy making.