Efficient Algorithms for Machine Learning

提议的研究是在“机器学习”领域，重点是开发具有实践和理论意义的高效算法。所研究的问题包括：(a)在存在噪声的情况下学习（在存在分类噪声的情况下搜索学习线性阈值和宇称函数，并且在存在分类噪声的情况下可能实现“假设增强”）；(b)有障碍物的探索（开发有效的算法，用于探索有障碍物和/或平面图的平面区域，以期将这些算法集成到更大的导航系统中）；(c)协调专家/特征选择（针对给定问题，制定有效的方法，将各种专家/特征的建议结合起来）；(d)因果结构的自主学习（探索主动学习如何有效地学习复杂环境，如Macintosh窗口系统）；(e)主动学习的新模式（改进对“成员查询”在学习中的作用和局限性的理解）；(f)学习和生物学（研究由分子生物学问题引发的各种推理问题）；(g)以不同的学习模式进行教学（重点关注“有效教学”的算法）。

The proposed research is in the area of ``machine learning,` with an emphasis on the development of efficient algorithms of both practical and theoretical interest. The problems investigated include: (a) learning in the presence of noise (searching for learning linear-threshold and parity functions in the presence of classification noise, and `hypothesis boosting` is possible in the presence of classification noise); (b) exploration with obstacles (developing efficient algorithms for exploring planar regions with obstacles and/or planar graphs, with a view towards integrating these algorithms in larger navigation systems); (c) coordinating experts/feature selection (developing effective methods of combining the advice of a variety of experts/features for a given problem); (d) autonomous learning of causal structures (exploring how active learning can effectively learn a complex environment, such as a Macintosh window system); (e) new models of active learning (refining the understanding of the power and limitations of `membership queries` for learning); (f) learning and biology (looking at a variety of inference problems motivated by problems in molecular biology); and (g) teaching in different learning models (focusing on algorithms for `effective teaching`).