The Thing We Tried That Worked: Utile Distinctions for Relational Reinforcement Learning
msra(2008)
摘要
This paper introduces a relational function approximation technique based on McCal- lum's UTree algorithm (McCallum, 1995). We have extended the original approach to handle relational observations using an at- tribute graph of observable objects and rela- tionships (McGovern et al., 2003). Further- more, we address the inherent challenges that arise with a relational representation. We use stochastic sampling to manage the search space (Srinivasan, 1999), and sampling to ad- dress issues of autocorrelation (Jensen and Neville, 2002). We prevent the algorithm from growing an overly large and complex tree by incorporating Iterative Tree Induc- tion's approach (Utgo, 1995). We compare Relational UTree's performance with similar relational learning methods (Finney et al., 2002) (Driessens et al., 2001).
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络