Learning to Catch: Applying Nearest Neighbor Algorithms to Dynamic Control Tasks
Learning to Catch: Applying Nearest Neighbor Algorithms to Dynamic Control Tasks
复制标题
学习捕捉：将最近邻算法应用于动态控制任务
DOI：
10.1007/978-1-4612-2660-4_33
复制
发表时间：
1994
期刊：
影响因子：
--
通讯作者：
S. Salzberg
中科院分区：
文献类型：
--
作者：
D. Aha;S. Salzberg
本文检验了 k 最近邻算法的局部加权变体可以支持动态控制任务的假设。我们在接住飞球的模拟学习任务中评估了几种 k-近邻 (k-NN) 算法。此前，针对此类问题已提倡使用局部回归算法。这些算法是 k-NN 的变体，它们的预测基于从最近邻计算的（可能是加权的）回归。虽然它们在许多任务上都优于 simplerk-NN 算法，但它们在接球任务上却遇到了麻烦。我们假设该任务中的非线性是造成这种行为的原因，并且可能需要修改局部回归算法才能在类似条件下正常工作。
This paper examines the hypothesis that local weighted variants ofk-nearest neighbor algorithms can support dynamic control tasks. We evaluated severalk-nearest neighbor (k-NN) algorithms on the simulated learning task of catching a flying ball. Previously, local regression algorithms have been advocated for this class of problems. These algorithms, which are variants ofk-NN, base their predictions on a (possibly weighted) regression computed from theknearest neighbors. While they outperform simplerk-NN algorithms on many tasks, they have trouble on this ball-catching task. We hypothesize that the non-linearities in this task are the cause of this behavior, and that local regression algorithms may need to be modified to work well under similar conditions.