Smoothness, Disagreement Coefficient, and the Label Complexity of Agnostic Active Learning
Smoothness, Disagreement Coefficient, and the Label Complexity of Agnostic Active Learning
复制标题
不可知主动学习的平滑度、不一致系数和标签复杂度
DOI：
10.5555/1953048.2021072
复制
发表时间：
2011-02
影响因子：
6
通讯作者：
Wang, Liwei
中科院分区：
文献类型：
--
作者：
Wang, Liwei
我们在存在噪声的情况下（即不可知环境）研究基于池的主动学习。众所周知，不可知主动学习的有效性取决于学习问题和假设空间。尽管主动学习在很多情况下非常有用，但也很容易构造出主动学习算法无法具有优势的示例。先前的工作表明，主动学习的标签复杂性依赖于分歧系数，而分歧系数往往表征学习问题的内在难度。在本文中，我们研究分类边界平滑且数据分布的密度可由平滑函数界定的分类问题的不一致系数。我们证明了有限和无限光滑问题的不一致系数的上限和下界。结合现有结果表明，对于平滑问题，主动学习优于被动监督学习。
We study pool-based active learning in the presence of noise, that is, the agnostic setting. It is known that the effectiveness of agnostic active learning depends on the learning problem and the hypothesis space. Although there are many cases on which active learning is very useful, it is also easy to construct examples that no active learning algorithm can have an advantage. Previous works have shown that the label complexity of active learning relies on the disagreement coefficient which often characterizes the intrinsic difficulty of the learning problem. In this paper, we study the disagreement coefficient of classification problems for which the classification boundary is smooth and the data distribution has a density that can be bounded by a smooth function. We prove upper and lower bounds for the disagreement coefficients of both finitely and infinitely smooth problems. Combining with existing results, it shows that active learning is superior to passive supervised learning for smooth problems.
登录
查看更多内容
影响因子：
2.5
作者：
R. Castro;R. Nowak
通讯作者：
R. Castro;R. Nowak
DOI：
10.1007/978-1-4757-2545-2
发表时间：
1996-03
期刊：
--
影响因子：
--
作者：
T. Mikosch;A. Vaart;J. Wellner
通讯作者：
T. Mikosch;A. Vaart;J. Wellner
DOI：
10.1145/1143844.1143853
发表时间：
2006-06
期刊：
Proceedings of the 23rd international conference on Machine learning
影响因子：
--
作者：
Maria-Florina Balcan;A. Beygelzimer;J. Langford
通讯作者：
Maria-Florina Balcan;A. Beygelzimer;J. Langford
影响因子：
4.5
作者：
Hanneke, Steve
通讯作者：
Hanneke, Steve
DOI：
--
发表时间：
2007-12
期刊：
--
影响因子：
--
作者：
S. Dasgupta;Daniel J. Hsu;C. Monteleoni
通讯作者：
S. Dasgupta;Daniel J. Hsu;C. Monteleoni