Data Sciences, AI and Machine Learning for Army Applications
Army Research Lab (ARL)
This project is focused on research in several disciplines from material science to data science and statistics to support the missions of the Aviation & Missile Center Technology Development Directorate. Our aim is to provide sound statistical and data science methodologies and data analysis for their research development and applications. The project is a two-year grant totaling $854,000 awarded to PI Prof. Rundensteiner, and Co-PIs Profs. Zou, Emdad, Zhang, and Kong.
Award Period: 2019-2021 Award Amount: $854,000
NRT-HDR: Data Driven Sustainable Engineering for a Circular Economy
National Science Foundation
This project is focused on training graduate students in disciplines from chemical science to data sciences to advance and support the future of circular economies. Our aim is to produce students versed in data-driven sustainable engineering that can have an impact on society. The project is a five-year traineeship grant totaling $2,999,289 awarded to PI Prof. Rundensteiner, and Co-Pis Profs. Paffenroth, Titova, Timko, and Deskins. For more information on this award, please visit https://www.nsf.gov/awardsearch/showAward?AWD_ID=2021871
Award Period: 9/1/2020 - 8/31/2025 Award Amount: $2,999,289
Modeling the dynamics of spindle behavior in cells with supernumerary centrosomes
NIH (R01 GM140465-01)
In Professor Olson's own words: Mitosis is the process of cell division, involving an intricate balance of forces to ensure a successful result—two genetically identical daughter cells. In normal cells, the mitotic spindle contains two spindle poles (bipolar), each having microtubules nucleated from a centrosome. Cells in disease states may have extra centrosomes, leading to either formation of a multipolar spindle and multiple daughter cells with poor viability, or formation of a pseudo-bipolar spindle with daughter cells that are viable. A hallmark of cancer cells is the ability to successfully divide with extra centrosomes. Through a combination of live-cell imaging and model simulations, we will provide new fundamental knowledge and insight into how the normal mitotic machinery has been co-opted to allow for bipolar division in cells with extra centrosomes. The developed modeling frameworks for fluid-structure interactions will lead to new computational methods that will leverage high performance computing architectures to simulate centrosome movement and stochastic MT dynamics.
Award Period: 9/5/2020 - 6/30/2023 Award Amount: $916,956
Valid time-series analyses of satellite data to obtain statistical inference about spatiotemporal trends at global scales
NASA/University of Wisconsin-Madison
Start Date: 2/21/2020 (2 years). Amount: $174,762.00
As remote sensing has matured, there is a growing number of datasets that have both broad spatial extent and repeated observations over decades. These datasets provide unprecedented ability to detect broad-scale changes in the world through time and to forecast changes into the future. However, rigorously testing for patterns in these datasets, and confidently making forecasts, require a solid statistical foundation that is currently lacking. The challenge presented by remotely sensed data is the same as its remarkable value: remotely sensed datasets consist of potentially millions of time series that are non-randomly distributed in space. We propose to develop new statistical tools to analyze big, remotely sensed datasets that will add rigor to the conclusions about patterns of past changes and confidence to forecasts of future trends. Our focus is providing statistical tests for regional scale hypotheses using pixel-scale data, thereby harnessing the statistical power contained within all of the information in remotely sensed time series.
Sarah Olson wins 2019 Mathematical Medicine & Biology Best Paper Prize
The winning paper was co-authored with Lucia Carichino (former WPI postdoctoral Scholar, now tenure track at RIT) and is titled "Emergent three-dimensional sperm motility: coupling calcium dynamics and preferred curvature in a Kirchoff rod model".
The journal Mathematical Medicine and Biology is managed by the Institute of Mathematics and its Applications (IMA) https://ima.org.uk, which is UK's chartered professional body for mathematicians and one of the UK's learned societies for mathematics. This journal publishes original articles with a significant mathematical content addressing topics in medicine and biology. It gives particular emphasis to papers exploiting modern developments in applied mathematics.
This best paper prize recognizes the most outstanding research published in the journal over a two year period, judged by a committee drawn from the Editorial Board.
In Sarah's own words: This paper extended previous modeling techniques to examine fully three-dimensional sperm motility where internal force generation is coupled to the relevant biochemistry. We focused on understanding the role of calcium coupling on swimming patterns; emergent trajectories that can be characterized as a hypotrochoid are observed for quasi-planar beatforms, similar to experiments
Portable multiplexed chemical agent sensor for detection in obscurant-heavy environments
DTRA and CCDC-SC
Start Date: 2020 (3 years).
This project is focused on combining machine learning with chemical sensor arrays to reduce false alarm rates in challenging environments. The project leverages our groups recent work in applying machine learning techniques to problems from the physical sciences. The project is a multi-party effort between WPI, CCDC-SC, Seiksui Chemical Co., and UMass Amherst. The project is an up to $1.8 million award to the team, and up to $249,000 of that amount is expected to support our group’s work on the project over the next three years.
DISTRIBUTION STATEMENT A. Approved for public release: distribution unlimited.
DeepM&Mnet: A General Framework for Building Multiphysics & Multiscale Models using Neural Network Approximation of Functions, Functionals, and Nonlinear Operators
Brown University (Prime: Defense Advanced Research Projects Agency)
Start Date: June 15, 2020 (3 years). Amount: $149,854
The proposal focuses on mathematical analysis for the neural network approximation (NNA) as well as the design of efficient deep learning algorithms for multiphysics and multiscale systems. The key idea is to train NNA for separated systems offline and then combine all the trained networks to solve multiphysics systems online. The research will result in fast and reliable solvers for multiphysics and multiscale systems. The main goal of the WPI investigator is to analyze and to improve the performance of the NNA as universal approximations in learning functions, functionals, and nonlinear operators and fusion of all pre-trained networks to solve multiphysics systems.The project is collaborative with Brown University and John Hopkins University
Simulating Large-Scale Morphogenesis in Planar Tissues
National Science Foundation, DMS 2012330
Start Date: 6/15/2020 (3 years). Amount: $200,000
Cutting-edge developments in biotechnology and medicine involve reconstructing large-scale tissues and organs. This work can be limited by lack of knowledge in tissue morphogenesis, the process by which living tissues develop their size-and-shape characteristics. Though live-imaging techniques have enabled the observation of morphogenetic processes, progress in fundamental understanding has been slow. This project aims to improve tools for modeling a wide range of living tissues that are relatively planar and have been extensively studied experimentally. The project will develop methods for numerical simulation of morphogenesis processes and attempt to reproduce the observed large-scale morphogenesis structures in planar tissues. The project provides graduate student training through involvement in the research.
This project concerns numerical simulation of large-scale continuum models for tissue morphogenesis that involve free boundaries, bulk-interface coupling, and highly nonlinear interactions. The work centers on a new mathematical model in which the field variables are nonlinearly coupled via reaction-convection equations and non-standard spatial partial differential equations. The project will develop semi-implicit and fully implicit time-stepping methods to avoid a potential time-step restriction for explicit time-stepping methods. Due to the high nonlinearity of the system, the boundary configuration must be updated together with the velocity field as well as other field variables. For this purpose, a novel interface-tracking method based on reference-map techniques will be investigated. Linear analysis close to trivial solutions will be conducted to assist the design of fast-converging iterative methods for solving the nonlinear system derived from the implicit time-stepping discretization of the original model. Simulations to understand in vitro micro-tissue and in vivo epithelial-tissue morphogenesis from live-imaging data will be carried out.
For more information on this grant, please visit https://www.nsf.gov/awardsearch/showAward?AWD_ID=2012330&HistoricalAward...