When: Monday, October 21, 2019
Where: PGH 563
Time: 11:00 AM
Focus on Research (FoR) is an opportunity for any COSC Ph.D. student to discuss a research project (with or without preliminary results), a conference dry run, or any research topic of interest to present to an audience of peers and faculty. It is a great avenue for PhD students to practice presentation skills in front of a larger and broader audience.
Enhancing Magnetic Resonance Imaging by Utilizing Deep Learning and Compressed Sensing Techniques
Nazanin Beheshti, Ph.D. Student
Magnetic Resonance Imaging (MRI) is a non-invasive and widely used imaging technique providing both functional and anatomical information for clinical diagnosis. However, long scanning and waiting time may result in motion artifacts and uncomfortable situation for patients. All these facts imply that speeding up the process of reconstructing image for MRI is a fundamental challenge.
Accelerating MRI by taking fewer measurements has the potential to reduce medical costs, minimize stress to patients and make MRI possible in applications where it is currently prohibitively slow or expensive. It has been shown that compressed sensing has had a great contribution in accelerating MRI process. However, compressed sensing technique does not always result in a reliable image with good quality.
Our goal is to combine compressed sensing and deep learning technique in k-space domain to acquire data without redundancy and apply a data dependent sparse representation in compressed sensing to reconstruct image.
Nazanin Beheshti is a 3rd year Ph.D. student. She is working under the supervision of Dr. Lennart Johnsson. Her research area is about high-performance computing and computer architecture. Her research is focused on designing high-performance and power-efficient solutions for challenges in memory and computation side of different algorithms for different systems.
Using Compression in MPI-IO: Semantics, Implementation, and Evaluation
Siddhesh Pratap Singh, Ph.D. Student
With an increase in data output due to the use of High Performance Computing (HPC) applications in big-data related fields, compounded by the comparatively slow access rates of hard drives, many parallel programs are bottlenecked by their I/O components. Other than performance degradation, this can also result in a decrease in disk space availability. A possible solution to this problem is compressing the data before it is written and decompressing it after being read. Writing in a compressed format will significantly reduce the space needed to store the file (which has many other benefits especially on a networked file system) and possibly reduce the time needed to read/write it (due to bandwidth limitations). OpenMPI is a Message Passing Interface (MPI) specification implementation which is widely used to develop applications for cluster computing. The functionality of OpenMPI is divided into components, each of which handles a specific task. Our aim is to create an OpenMPI I/O component which compresses and decompresses I/O using the Snappy library with its read and write functions to improve I/O performance for large files.
Siddhesh Pratap Singh is a 3rd year Ph.D. student in computer science at The University of Houston studying under the guidance of Dr. Edgar Gabriel. His area of research is high performance computing specifically with respect to OpenMPI.