Accuracy-Aware Optimization of Approximate Programs
|Event Date:||March 25, 2015|
|Speaker Affiliation:||Massachusetts Institute of Technology|
|Contact Name:||Professor Milind Kulkarni
Many modern applications (such as multimedia processing, machine learning, and big-data analytics) exhibit a natural tradeoff between the accuracy of the results they produce and the application's execution time or energy consumption. These applications allow us to investigate new optimization approaches that are beyond the reach of standard program optimizations.
I present a novel approximate optimization framework based on accuracy-aware program transformations. These transformations trade accuracy in return for improved performance, energy efficiency, and/or resilience. The optimization framework includes program analyses that characterize the accuracy of transformed programs, and search techniques that navigate the tradeoff space induced by transformations to find approximate programs with profitable tradeoffs. This talk will show how we can use this accuracy-aware optimization framework to 1) automatically generate approximate programs with significantly improved performance and acceptable accuracy, and 2) automatically generate approximate functions that maximize energy savings when executed on approximate hardware platforms, while ensuring that the generated functions satisfy the developer's accuracy specifications.
Sasa Misailovic is a Ph.D. Candidate in Electrical Engineering and Computer Science at MIT. His interests include programming languages, software engineering, and computer systems, with an emphasis on improving performance, energy efficiency, and resilience in the face of software errors and approximation opportunities. His recent research on systems for analyzing and optimizing programs that execute on approximate hardware received two best paper awards at OOPSLA 2013 and OOPSLA 2014. (http://people.csail.mit.edu/misailo/)