-
Notifications
You must be signed in to change notification settings - Fork 0
/
info.json
17 lines (17 loc) · 1.72 KB
/
info.json
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
{
"abstract": "Beam search is commonly used to help maintain tractability\nin large search spaces at the expense of completeness and optimality. Here we\nstudy supervised learning of linear ranking functions for controlling\nbeam search. The goal is to learn ranking functions that allow for beam search to\nperform nearly as well as unconstrained search, and hence gain computational\nefficiency without seriously sacrificing optimality. In this paper, we develop\ntheoretical aspects of this learning problem and investigate the application of\nthis framework to learning in the context of automated planning. We first study\nthe computational complexity of the learning problem, showing that even for\nexponentially large search spaces the general consistency problem is in NP. We\nalso identify tractable and intractable subclasses of the learning problem,\ngiving insight into the problem structure. Next, we analyze the convergence\nof recently proposed and modified online learning algorithms, where we introduce\nseveral notions of problem margin that imply convergence for the various algorithms.\nFinally, we present empirical results in automated planning, where ranking\nfunctions are learned to guide beam search in a number of benchmark planning\ndomains. The results show that our approach is often able to outperform an existing\nstate-of-the-art planning heuristic as well as a recent approach to learning such\nheuristics.",
"authors": [
"Yuehua Xu",
"Alan Fern",
"Sungwook Yoon"
],
"id": "xu09c",
"issue": 54,
"pages": [
1571,
1610
],
"title": "Learning Linear Ranking Functions for Beam Search with Application to Planning",
"volume": "10",
"year": "2009"
}