Differences

seminar_09_29_09 [2009/10/25 23:36] yoonguk created		seminar_09_29_09 [2009/10/25 23:45] (current) yoonguk
Line 14:		Line 14:
	Search (using an execute and time feedback system) however, has several unfortunate consequences: it is time consuming (large search space); it is subject to measurement noise; it gives the best algorithm, rather than the optimal; it may not even find the best algorithm; and finally, it is a black box (we know which algorithm is best, but not why it is best).		Search (using an execute and time feedback system) however, has several unfortunate consequences: it is time consuming (large search space); it is subject to measurement noise; it gives the best algorithm, rather than the optimal; it may not even find the best algorithm; and finally, it is a black box (we know which algorithm is best, but not why it is best).
	The goal of this work is to build an optimization framework for linear transforms that relies on the parameters and structure of the memory hierarchy ra-ther than search. The approach involves the definition of a memory model on which we define a cost function. We then identify what structures in the transform algorithm minimize the cost function and construct DFT algorithms that meet these crite-ria.  This talk describes on-going research and many results are preliminary.		The goal of this work is to build an optimization framework for linear transforms that relies on the parameters and structure of the memory hierarchy ra-ther than search. The approach involves the definition of a memory model on which we define a cost function. We then identify what structures in the transform algorithm minimize the cost function and construct DFT algorithms that meet these crite-ria.  This talk describes on-going research and many results are preliminary.
-
	=====Bio=====		=====Bio=====
	Marek Telgarsky is a PhD student in the Electrical and Computer Engineering department. He com-pleted his Bachelor's Degree in Electrical Engineer-ing at the University of Wisconsin-Madison. His ad-visors are Jose Moura and James Hoe. His re-search interests include signal processing, software optimization, and computer architecture.		Marek Telgarsky is a PhD student in the Electrical and Computer Engineering department. He com-pleted his Bachelor's Degree in Electrical Engineer-ing at the University of Wisconsin-Madison. His ad-visors are Jose Moura and James Hoe. His re-search interests include signal processing, software optimization, and computer architecture.
		+	**[[Seminars | Back to the seminar page]]**