Mentors: Selin Aviyente, Abdullah Karaaslanli
In this project, the student will work with high dimensional networks depicting connectivity in the human brain. These networks are derived from electrical activity of the human brain, i.e. electroencephalogram (EEG). The student will work on developing clustering and community detection algorithms for determining the structure of these networks and how they change across time and frequency.
By the end of this project students will implement different community detection methods; analyze graphical data and learn how to use different graph visualization tools
Students applying to this project will mostly be using MATLAB or Python. Some background in linear algebra would be very beneficial.
Keywords: Machine learning & Neuroscience
Mentors: Dirk Colbry
The SEE-Insight Team is developing image understanding tools which focus on common workflows used in scientific image understanding for applications in engineering, medical imaging, biology, etc. The goal is to develop tools that are more than just a manual annotation system. As the scientist annotates their images, the tools will take their annotations (starting from the very first image) and use machine learning to search the “algorithm space” to try and identify candidate algorithms based on their specified workflow. If a good candidate algorithm is found, then the tools will start making suggestions to the researcher. In the worst case, using just the tools will take no more time than scientists would need to annotate the images manually; the result will be an annotated dataset which they can use to conduct their science or to feed into a more traditional ML system. However, in the best case, a good candidate algorithm can be automatically identified, that can help speed up some, or all, of the annotation process, saving researcher time and speeding up their overall research workflow.
By the end of this project students will learn about developing researcher/user interfaces, using artificial intelligence and scientific image understanding. Our group uses best practices in scientific software development using tools such as GIT and Python. We also have opportunities to learn to use large scale computing systems such as the MSU high performance computing center, Open Science Grid, XSEDE resources and Cloud based resources (AWS, Google Cloud, Azure etc).
Students applying to this project will mostly be using Python but also a little Node.js for user interface design. Students from a wide range of backgrounds are encouraged to apply.
Keywords: Machine learning and scientific image understanding
Mentors: Alex Dickson, Kin Sing Steven Lee and Tom Dixon
When drugs are administered, they undergo processes such as adsorption, distribution, metabolism and excretion (commonly referred to as ADME). These can be described by a set of ordinary differential equations, allowing us to predict the time course of drug action, including the length of time it remains in the body. In this project we will use computational modeling to investigate how the microscopic properties of a drug molecule affect its efficacy.
By the end of this project students will be able to model systems of ODEs, connect models to experimental data, ask and answer precise research questions, program in Python, and work with code repositories.
Students applying to this project should be able to program in python.
Keywords: drug design, python, ordinary differential equations
Mentors: Arjun Krishnan
Gene expression profiling is a powerful technique for recording the activity of all 25,000 genes in the genome in any human tissue sample. The goal of this project is to develop machine learning approaches that analyze >200,000 gene-expression profiles to: i) unearth signals that are highly indicative of the disease status of the individual who gave the sample and, ii) predict potential combinations of drugs that can reverse the disease-related signals, by integrating drug-response data.
By the end of this project, students will be able to: Download, process, and explore complex, heterogeneous biomedical datasets; Build supervised and unsupervised machine learning approaches to mine these datasets; Make powerful visualizations that summarize data and research findings; and Effectively communicate the entire project to both biologists and computational scientists.
Students applying to this project should have some familiarity with data-wrangling and visualization in Python (with NumPy, Pandas, Matplotlib/Seaborn). Familiarity with machine learning toolkits in Python (Scikit-Learn) is a bonus.
Keywords: Computational Genomics, Biomedical Data Science, Machine Learning, Bioinformatics
Anthropogenic water management, Climate Change, and Environmental Sustainability in the Southwestern US
Mentors: Lifeng Luo, Yadu Pokhrel
The Southwestern US is facing significant water sustainability challenges due to climate change and human water demands. This project will use a modeling framework that is based on the Community Land Model version 5 (CLM5) combined with the river-floodplain-reservoir routing model called the CaMa-Flood. The modeling system also simulates irrigation, groundwater pumping, and environmental flow requirements. Simulations will be carried out for both historical periods and future projections under different climate and socioeconomic scenarios to explore solutions for water sustainability in the region.
By the end of this project, students will be able to work with a state-of-the-art land surface modeling system to simulate the hydrological cycle in a watershed or basin. Students will have a better understanding about how to model natural hydroloigcal processes and human water management within the hydroloigcal cycle. Students will develop skills in Fortran programming and parallel computing.
Students applying to this project should be able to program using python, Fortran and bash. Additionally, a basic understanding about hydrology is required and some exposure to computational hydrology would be helpful but not required.
Keywords: hydrological and climate modeling
Mentors: Liz Munch, Firas Khasawneh, Melih Yesilli
Signal processing is the field of research dedicated to understanding and predicting behavior of a dynamical system given a 1D signal. Topological signal processing seeks to understand the signal by bringing tools from topological data analysis (TDA), which has tools for quantifying shape in data. In this project, depending on student interest, we will investigate some of the available tools, how they can be combined with machine learning approaches, and how they can be applied to data sets such as seismic data from earthquakes.
At the end of this project, students will understand some of the available tools from topological signal processing, such as the delay coordinate embedding, network methods for signal processing, and persistent homology approaches to their analysis.
Students applying to this project will use python and should have previously completed coursework in linear algebra.
Keywords: topological signal processing
Mentors: Saiprasad Ravishankar, Haiyan Huang
The project will focus on models and machine learning-based algorithms for computational imaging, with potential theoretical/mathematical work as well. The students will work with the SLIM group and Dr. Ravishankar in developing and implementing methods particularly involving novel machine learning for image formation and image processing in medical, scientific, and industrial imaging applications. The project may also involve interactions with Dr. Ravishankar's collaborators in universities, national labs, and industry.
By the end of this project, students would have gained some familiarity with the field of computational imaging and algorithms and machine learning in that domain. The students would have taken part in and contributed to active research projects, potentially leading to a publication.
Students applying to this project will primarily use Python and Matlab. Additional courses that would be recommended or very helpful for research with the SLIM group would be linear algebra, optimization, and signal processing in that order of importance.
Keywords: Machine learning for imaging
Mentors: Devin Silva, Danny Caballero, Tom Finzell and Rachel Frisbie
Michigan State undergraduates from a broad range of backgrounds take our introductory computational science course sequence, but there is little research available on how students learn to do computational coursework outside of traditional computer science courses. We plan to assess their learning outcomes and attitudes toward computational science coursework after they have completed these courses. Our summer research student(s) will conduct interviews with former CMSE 201/202 students to gain insight into their experiences. After designing and conducting the interviews, students will work on analyzing and “coding” the data to identify important themes related to student understanding of computational topics and student attitudes toward such topics.
By the end of this project, students will be able to design and conduct education research interviews and perform qualitative data analysis.
Students applying to this project will code in python.
Keywords: Computational education research
Uncovering hidden signals in satellite imagery to improve disturbance detection and biodiversity models
Mentors: Phoebe Zarnetske, Jasper Van doninck
This research will provide key data to quantify disturbances from satellite imagery across local to continental scales, and to improve models of biodiversity response to environmental changes in space and time. Current satellite image time series analysis methods typically provide a simplified description of the spectral behavior of a single pixel through time. This focus on individual pixels means that potentially valuable spatial information is lost: both the temporal features of interest (e.g. land cover change) as noise (e.g. cloud cover) have spatial patterns that could provide additional information. In this project you will develop code to perform spatial segmentation of satellite image time series based on temporal correlations between adjacent pixels. Spatial segments can then be used in a time series analysis instead of individual pixels, advancing the ability to quantify disturbance and provide robust data for biodiversity modeling.
By the end of this project, students will be able to write code to manipulate large image time series datasets, learn spatial segmentation algorithms, increase their programming skills in R and Google Earth Engine, understand the linkages between disturbance and biodiversity, and contribute to creating high resolution disturbance spatial data for the scientific community.
Keywords: Remote sensing image analysis