Categories: General
      Date: Feb  7, 2017
     Title: Muzahid receives $450K NSF CAREER Award


 

Abdullah Muzahid, assistant professor in the UTSA Department of Computer Science, has been awarded a National Science Foundation (NSF) Faculty Early Career Development (CAREER) Award. The award includes a $450,000 grant to support his research in machine learning over the next five years. He is the seventh faculty member in Computer Science to receive this award.


His project, CAREER: A Dynamic Program Monitoring Framework Using Neural Network Hardware, focuses on utilizing a specialized hardware that implements neural network. Neural network is a machine learning technique that mimics human brain. Therefore, neural network hardware provides some unique capabilities that can be utilized in many different ways.

 

The usage that this project considers is about "program monitoring." Program execution monitoring is often used to detect software bugs, performance issues, security attacks etc. When neural network hardware is used to monitor program executions, at first it tries to learn the normal behavior of the program. Such behavior is defined in terms of various events or features that the program exhibits.

 

Once the neural network hardware learns the normal behavior sufficiently, it looks for any deviation of that behavior. Such deviation can be attributed to software bugs, performance issues or security attacks. By focusing on different behavior, the proposed approach can detect different software bugs, performance issues or security attacks. Since the neural network hardware can learn any new behavior, this approach is adaptive, can be done on-the-fly, and creates a general framework for bugs and security attacks. The project proposes a multi layer approach consisting of hardware, runtime system, and compiler to implement this idea.

 

Besides this project, Dr. Muzahid is leading few other projects. One project is exploring the idea of approximate computing. There is a domain of applications (such as scientific simulation, modeling, graphics etc.) that does not require 100% accurate results. For such applications, we can introduce some inaccuracy (e.g., by relaxing certain constraints on synchronization, communication etc.) and still be able to produce acceptable results. On the positive side, the relaxation of various constraints can enable simplification of hardware, runtime system or even programming models and help us improve overall performance and energy efficiency. Other projects look into issues ranging from multiprocessor's memory model, data center scheduling, and workload characterization.

 

Besides the NSF CAREER award, Dr. Muzahid was previously awarded another NSF grant of around $250K. Currently, he is in the process of receiving a seed grant from Intel Corporation.