Meet our 2018–2019 Postdoctoral Fellows!
Lara Reichmann and Abbie Popa joined the Data Institute at the University of San Francisco as postdoctoral fellows following academic careers in Ecology and Evolutionary Biology, and Neuroscience, respectively. Continue reading to learn about their interests and experiences at the Data Institute!
What did you get your Ph.D. in?
ABBIE — I got my Ph.D. in neuroscience at UC Davis. My primary project looked at electrical signals from the brain (EEG) associated with attention and behavior control. I compared how adolescents with and without anxiety responded to tasks around different types of images. In that study, I found that adolescents with anxiety had more trouble controlling their attention to mildly threatening images, but were good at withholding an action in response to mildly threatening images.
LARA — I got my Ph.D. in Ecology and Evolutionary Biology from Brown University. My doctoral dissertation was in the field of ecosystem ecology, studying the effects of changing precipitation on dryland ecosystems’ primary productivity, i.e., the rate at which plants fix carbon, and on the cycling of nitrogen. I carried out manipulative experiments in the Chihuahan desert, and in the grasslands of Texas and California. So, in general, I spent a lot of time gathering data in the field!
How did you develop an interest in data science?
ABBIE — When I first started in my Ph.D. lab I found that I had a lot of data points from a lot of measures, and it was suggested to me that I could better understand my sample if I looked at these measures organized with k-means clustering. I was pleased with how this data-driven technique gave me a better understanding of my data.
Additionally, most of the methods we use in cognitive neuroscience produce a lot of data points. For example, EEG and eye-tracking produce a measure every 2–8 milliseconds, so even if your experiment is only 5 minutes long you swiftly end up with a lot of data! MRI, similarly, produces a lot of data because you look for signal breaking the brain down into millions of voxels. Additionally, brain data often have underlying correlations and other structure that you have to account for during processing and analysis. I really appreciated how data science techniques provided me with a way to find the story in this type of large, messy data.
I joined a student directed study group (Davis Incubator Group) to help practice my data science skills, and later when UC Davis started the Data Science Initiative I became an affiliate there. These were both really fun for me, because I got to apply data science problem solving to lots of different domains in addition to neuroscience.
LARA — My interest in data science grew during my Ph.D. studies, especially when I worked at the Jornada, a research site in New Mexico within the LTER Network (which stands for Long Term Ecological Research). The LTER Network was founded in 1980 with the idea that long-term research could help unravel patterns in nature that would not be apparent with short-term studies alone. Data from the Network opens the opportunity for scientists around the world to test hypotheses across different ecosystems and times. Data science tools make it possible to make the most out of the Network’s data.
Later, during my postdoc at the USDA, I conducted meta-analyses of published data to understand the effect of increased precipitation variability on crops that could be used for biofuel production. Predictions on climate change’s effects on ecosystems usually focus on changes in the mean (for example, increases in average temperature or increases in the average amount of rain), but these predictions rarely tell us what could happen as extreme climatic event become more frequent. At the USDA, I used long-term climate data to link a site’s precipitation variability with switchgrass yields. A programming language like R was essential in analyzing this data.
Why did you choose to do your postdoctoral fellowship with the Data Institute at the University of San Francisco?
ABBIE — I knew I wanted to continue exploring parts of data science I didn’t get as much exposure to at UC Davis. The USF data science postdoctoral fellowship was unique in that it offers opportunities to both do more traditional academic research and collaborate with private sector companies. The faculty here are also expert in a lot of cutting age data science methods like deep learning, which gives me the opportunity to apply these methods to data from my neuroscience roots!
LARA — The Date Institute provides a unique opportunity to acquire a new set of tools to broaden the way I do science, and to interact with other scientists and researchers outside the field of environmental science. The Data Institute is also quite appealing because it partners with nonprofits and industries; settings very different from my experiences in the laboratory. This has exposed me to the latest trends and applications of data science.
Tell us about your current research as a postdoctoral fellow with the Data Institute.
ABBIE — I have started two research projects. For one I am working with Dr. James Wilson, who is a professor at the Data Institute and an expert in network analyses. For this project, we are looking at functional networks generated from MRI data in individuals with and without schizophrenia. We are using an approach that generates feature embeddings for each of the nodes in the brain, similar to how word2vec feature embeddings describe and organize words. Then we can use these feature embeddings to examine the organization of the functional brain networks in the two groups.
For the second project I am working with Dr. Bill Bosl, who is a professor in the School of Nursing and is affiliated with the Data Institute. Bill has developed an approach to EEG data that extracts non-linear features such as entropy from the brain signals, providing more features to examine beyond the traditional power spectra measures. Previously, Bill published a paper using these features to predict the development of autism in infants. Currently, we’re working on using these features for prediction of clinical outcomes in infants who were born preterm. We’re also working on expanding and validating the feature engineering part of the machine learning pipeline.
LARA — I am currently involved in two very exciting research projects. In one, I am collaborating with Schmidt Marine Technology grantee Dr. Brian Glazer, a professor at the University of Hawai’i at Mānoa. My work helps his lab to create an affordable coastal observatory platform that will integrate ocean sensors operating in real time. I work side-by-side with Connor Swanson (MS Data Science at USF) and Dr. David Uminsky (USF Data Institute). We are developing the data throughput, the front-end data visualization tools, and the algorithms to expand and improve predictive models of ocean tides. What I like most about this project is that Dr. Glazer’s goal is to make environmental sensors more user-friendly. As a frequently-frustrated user of environmental sensors with cryptic interfaces, this really appeals to me!
My other project is a collaboration with Dr. Gilmer Valdes (UCSF), and Dr. Yannet Interian (USF Data Institute), in which we are developing neural network models to get a deeper understanding of the causes of death at hospital intensive care units (ICUs), and exploring bias in current predictive models. We are using the available data to compare the diagnostic ability and specificity of our predictive models to that of other predictive models. The goal of this project is to inform health care policies, help identify new areas of research or improvement in patient care, and hopefully save lives.
What is your favorite thing so far about being a postdoctoral fellow at the Data Institute?
ABBIE — The espresso machine! (Kidding.) Really, it’s a great community of people to be around for a postdoctoral fellowship. For one, the faculty are always eager to work with us and support our projects and career growth. I also really like being in a community of technically minded people who can support using the best tool for the job.
LARA — My favorite part of being a postdoctoral fellow at the Data Institute has been interacting with a very diverse and talented group of students and faculty. The breadth of expertise at the Data Institute creates a nourishing and encouraging environment for developing my research and teaching skills.
Have you taken advantage of the MS in data science course access and/or certificate courses?
ABBIE — I have been sitting in on some of the courses! This was another big advantage of this postdoc position for me. Often in traditional academic environments researchers don’t use the most sophisticated computing tools, so it has been really nice to add knowledge about SQL and Spark to my data science toolbox. I’ve also appreciated the evening coursework in deep learning. As a Ph.D. researcher I dabbled in deep learning, but I have been able to get a much more formal and thorough understanding of best practices through these courses.
LARA — I’ve certainly taken advantage of the courses and certificates. With the help of my faculty mentors, I have chosen classes to broaden my knowledge in data science, especially those in machine learning/deep learning. These classes have not only been fascinating, but they directly relate to my research interests.
Lara, how would you like to apply Data Science to Ecosystem Ecology?
This fellowship has showed me that the applications of Data Science to Ecology and Environmental Science are limitless. I am very interested in applying Data Science to the restoration and management of human-altered landscapes, and the mitigation of the effects of climate change. I think I could do this in various ways; from designing automation processes that extract data from photos (which is often a very tedious task, done by hand, and likely assigned to a new or novice lab member), to developing models to predict natural phenomena.
Abbie, how do you connect Neuroscience to Data Science?
I think the connection between neuroscience and data science goes both ways. I’m often struck by how things I learned in my Neuroscience coursework show up in data science. For example, I still have flashbacks about having to write out the convolutions for edge detection and corner detection during my neuroscience comprehensive exam every time they come up in deep learning classes. On the other hand, neuroscience has many techniques, like functional MRI, that are young enough that researchers are still figuring out how to best parse and interpret the data. I really believe data science can inform how to dig into this data, since data science has found ways to cope with a lot of the elements of these data sets like size and network structure.
For more information on the Data Institute’s postdoctoral fellowship:
