Automatic Feature Generation for Machine Learning for Compilers
|Starts:||14:00 29 Feb 2012|
|Ends:||15:30 29 Feb 2012|
|What is it:||Seminar|
Speaker: Dr Hugh Leather. University of Edinburgh
Host: Alasdair Rawsthorne
Machine learning has been shown to automate and in some cases outperform hand crafted compiler optimizations. Central to such an approach is that machine learning techniques typically rely upon summaries or features of the program. The quality of these features is critical to the accuracy of the resulting machine learned algorithm; no machine learning method will work well with poorly chosen features. However, due to the size and complexity of programs, theoretically there are an infinite number of potential features to choose from. The compiler writer now has to expend effort in choosing the best features from this space.
The main part of this talk describes the work I have done attack this problem, developing a novel mechanism to automatically find those features which most improve the quality of the machine learned heuristic. The feature space is described by a grammar and is then searched with genetic programming and predictive modeling. This was the first work to consider the searching the feature space and the results will show considerable improvement over human derived features.
In the latter part of the talk I will describe some of the current projects I am working on.
Travel and Contact Information
Lecture Theatre 1.4