|
Activity Number:
|
31
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 2, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Learning and Data Mining
|
| Abstract - #303742 |
|
Title:
|
Efficient Large Design Space Exploration and Prediction in Computer Microarchitectural Study
|
|
Author(s):
|
Bin Li*+ and Lu Peng
|
|
Companies:
|
Louisiana State University and Louisiana State University
|
|
Address:
|
Room 61 Agriculture Administration Building, Baton Rouge, LA, 70803,
|
|
Keywords:
|
regression tree ; boosting ; active learning ; MART ; maximin sampling
|
|
Abstract:
|
Computer architects use cycle-by-cycle simulation to evaluate design choices and understand tradeoffs (between processor performance and power consumption) and interactions among design parameters. Efficiently exploring the exponential-size design spaces with many interacting parameters remains an open problem: the sheer number of experiments renders detailed simulation intractable. However, only configurations in a subspace can be simulated in practice due to long simulation time and limited resource, leading to suboptimal conclusions which might not be applied to unsampled design configurations. In this study, we propose an automated design space exploration and prediction method which employs sampling technique from experiment design and machine learning, and predictive modeling in data mining.
|
