Dr. Martin Luther King, Jr. Library: Applied Data Science: Overview

Welcome from Your Librarian

As your liaison librarian, I'm responsible for collaborating with ADS students and faculty for their research and educational needs. Please feel free to reach out to me anytime you have questions or suggestions regarding research or our resources. Please see my contact information in the profile box on this page and feel free email me or schedule a consult. Thank you!

Selected Publications

Automatic Modulation Classification of Frequency-Hopping Signals Using High-Dimensional Phase DiagramsThis link opens in a new window Automatic modulation recognition for frequency-hopping (FH) signals remains very challenging to researchers due to the signals' time-varying spectral characteristics. In this work, a novel robust automatic modulation recognition scheme is investigated for FH signals using the phase-space topological features represented by the embedded phase diagrams. As such embedded phase diagrams are often high-dimensional, it is necessary to formulate the phase-space features as tensors. In the training process, the phase-space tensor features will be utilized to establish the regression models as linear encoders for the individual modulations. The aforementioned linear encoders are constructed using the support vector machine (SVM); the phase-space feature-tensors of the training signals of all modulations will be projected by their corresponding regression models (or linearly encoded) to produce the representative code-vectors, respectively. In the test stage, the phase-space feature-tensor produced from a test signal will be projected by each individual trained regression model (or linearly encoded) to generate the respective code-vectors. Then, the code-vectors resulting from the test stage will be compared with the representative code-vectors to find which modulation will lead to the smallest Euclidean distance in between and such a modulation will be picked as the modulation type of the test signal. Monte Carlo simulation results have demonstrated that the average recognition accuracy of our proposed new approach is more than 90% when the signal-to-noise ratio is no less than 0 dB for additive white Gaussian noise. Jul 15, 2024
Distributed IoT Community Detection via Gromov-Wasserstein MetricThis link opens in a new window The Internet of Things (IoT) network is a complex system interconnected by different types of devices, e.g., sensors, smartphones, computers, etc.. Community detection is a critical component to understand and manage complex IoT networks. Although several community detection algorithms were proposed, they in general suffer several issues, such as lack of optimal solutions and scalability, and difficulty to be applied to a dynamic IoT environment. In this work, we propose a framework that uses Distributed Community Detection (DCD) algorithms based on Gromov-Wasserstein (GW) metric, namely GW-DCD, to support scalable community detection and address the issues with the existing community detection algorithms. The proposed GW-DCD applies Gromov-Wasserstein metric to detect communities of IoT devices embedded in a Euclidean space or in a graph space. GW-DCD is able to handle community detection problems in a dynamic IoT environment, utilizing translation/rotation invariance properties of the GW metric. In addition, distributed community detection approach and parallel matrix computations can be integrated into GW-DCD to shorten the execution time of GW-DCD. Finally, a new metric, i.e., Gromov-Wasserstein driven mutual information (GWMI), is derived to measure the performance of community detection by considering internal structure within each community. Numerical experiments for the proposed GW-DCD were conducted with simulated and real-world datasets. Compared to the existing community detection algorithms, the proposed GW-DCD can achieve a much better performance in terms of GWMI and the runtime. Jul 15, 2024
Learning black- and gray-box chemotactic PDEs/closures from agent based Monte Carlo simulation dataThis link opens in a new window We propose a machine learning framework for the data-driven discovery of macroscopic chemotactic Partial Differential Equations (PDEs)—and the closures that lead to them- from high-fidelity, individual-based stochastic simulations of Escherichia coli bacterial motility. The fine scale, chemomechanical, hybrid (continuum—Monte Carlo) simulation model embodies the underlying biophysics, and its parameters are informed from experimental observations of individual cells. Using a parsimonious set of collective observables, we learn effective, coarse-grained “Keller–Segel class” chemotactic PDEs using machine learning regressors: (a) (shallow) feedforward neural networks and (b) Gaussian Processes. The learned laws can be black-box (when no prior knowledge about the PDE law structure is assumed) or gray-box when parts of the equation (e.g. the pure diffusion part) is known and “hardwired” in the regression process. More importantly, we discuss data-driven corrections (both additive and functional), to analytically known, approximate closures. Jul 15, 2024
aBnormal motION capture In aCute Stroke (BIONICS): A Low-Cost Tele-Evaluation Tool for Automated Assessment of Upper Extremity Function in Stroke PatientsThis link opens in a new window Background: The incidence of stroke and stroke-related hemiparesis has been steadily increasing and is projected to become a serious social, financial, and physical burden on the aging population. Limited access to outpatient rehabilitation for these stroke survivors further deepens the healthcare issue and estranges the stroke patient demographic in rural areas. However, new advances in motion detection deep learning enable the use of handheld smartphone cameras for body tracking, offering unparalleled levels of accessibility. Methods: In this study we want to develop an automated method for evaluation of a shortened variant of the Fugl-Meyer assessment, the standard stroke rehabilitation scale describing upper extremity motor function. We pair this technology with a series of machine learning models, including different neural network structures and an eXtreme Gradient Boosting model, to score 16 of 33 (49%) Fugl-Meyer item activities. Results: In this observational study, 45 acute stroke patients completed at least 1 recorded Fugl-Meyer assessment for the training of the auto-scorers, which yielded average accuracies ranging from 78.1% to 82.7% item-wise. Conclusion: In this study, an automated method was developed for the evaluation of a shortened variant of the Fugl-Meyer assessment, the standard stroke rehabilitation scale describing upper extremity motor function. This novel method is demonstrated with potential to conduct telehealth rehabilitation evaluations and assessments with accuracy and availability. Jul 15, 2024
Physics-informed Machine Learning Models for Go/No-go Criteria on Reactive MetamaterialsThis link opens in a new window We present a physics-informed machine learning framework for predicting Go/No-Go criteria for reactive metamaterials and study shock propagation through a one-dimensional laminate structure. The laminate material was composed of an HMX bed with equally distributed 2mm thick copper pillars. The Wide-Ranging equation of state (WR EOS) was used to model HMX while the Romenski EOS was used for the elastic regime of copper, with the assumption of perfect plasticity. The shock was initiated by using an aluminum impactor and gauges were placed at the entry of the first copper pillar and exit of the last pillar. A modified machine learning model was then developed to predict the Go/No-Go criteria for the laminate structure. The proposed model only uses short-time measurements for predicting this behavior, that leads to large reductions in computational cost at higher dimensions. This framework suggests a data-driven guideline for the design of optimal laminate structures (e.g. number of copper pillars, thickness, and distribution). Jul 15, 2024

Website | Feed