Power prediction is a critical necessity as chip sizes continually decrease and the desire for low power consumption is a foremost design objective. For such predictions, it is crucial to avoid underestimating power since reliability issues and possible chip damage might occur. It becomes necessary to eliminate or strictly limit underestimations by relaxing accuracy constraints while decreasing the likelihood that the estimation undershoots the actual value. Also during the design stage, it is important to efficiently prune out to poor architecture configuration choices from the large microarchitectural design space which is usually tested using software simulations to determine the exact performance metric and the power consumption, a typically computationally expensive task. Motivated to propose some energy aware machine learning techniques for such unbalanced search spaces, we modified vanilla flavor SVM and GA algorithms both of which outperformed results achieved on the same datasets by previously published work. Aside from the microarchitectural design space project, Dr. Awad will be also presenting some of her research work on visual search, biologically inspired deep belief networks, wild sport video analysis, augmented reality and interactive environments using mobile apps .
- Acerca del IIIA