View Position
| Project Title | High Performance Data Analytics for Brain State Modeling |
| Summary | This project aims to build a system for dynamic and onsite brain state analysis using physiological data. Our existing system consists of a web server with an SQL database for data storage, mobile app with user interface, and EEG headset. The web server performs onsite data analysis and testing on selected data from the database. High performance computing platform provides computational power to ensure real time processing of requests. The student will work on feature extraction from physiological data and perform thorough testing to maximize the recognition rates of the brain state models. Parallel programs, including Parallel R, will be used to develop the back-end programs to ensure real-time processing the physiological data. |
| Job Description | The challenge of modeling brain waves is to design an inclusive system, which incorporates all the major brain waves together with other physiological data to build brain state models with satisfactory recognition rates. It is imperative for brain wave modeling studies to contemplate the rigorous time series analysis of brain waves to decipher trend, irregularities, cycles, seasonality and other variations among waves during different states. For example, existing study indicates that a model trained using data from a subject often does not classify data from a different subject correctly. Sometime data from the same subject collected at different times do not work well with each other, either. Therefore, feature extraction is an important part of EEG data analysis. The project aims to address the complexity of classification of brain waves data by modeling the major brain waves and achieve an efficient and predictable brain wave modeling system which has potential application in hospitality and clinical industry for self-controlled deep brain relaxation and early diagnosis of various brain abnormalities respectively. To ensure real-time processing of the physiological data and instant brain state classification, parallel programs will be used to perform dynamic machine learning algorithms. |
| Use of Blue Waters | Yes, we will use Blue Waters to develop and test parallel machine leaning programs. In particular, we will investigate how to use Parallel R on Blue Waters. UHD is also building its own cluster through a DOD grant. We will most probably test our programs locally first and then load onto Blue Waters. We will compare the performance of the parallel program to the corresponding sequential program and evaluate the speedup of the parallel program. We will summarize the results and assess the effectiveness of the parallel machine learning algorithms in the real-time dynamic brain state modeling process. |
| Conditions/Qualifications | Must understand machine leaning algorithms and be familiar to R. Understadning of SQL and PHP is a plus. |
| Start Date | 05/15/2017 |
| End Date | 05/14/2018 |
| Location | Department of Computer Science and Engineering Technology University of Houston-Downtown Houston, Texas 77002 |
| Interns | Caleb Ji
|