Information Criteria for Kernel Machines
Information Criteria for Kernel Machines
复制标题
DOI：
--
复制
发表时间：
2005
期刊：
影响因子：
--
通讯作者：
Kei Kobayashi;F. Komaki
中科院分区：
文献类型：
--
作者：
Kei Kobayashi;F. Komaki
提出了核正则化信息准则（KRIC），它是核逻辑回归（KLR）和支持向量机（SVMs）中调整正则化参数的新准则。KRIC的主要思想是基于正则化信息准则（RIC）。虽然RIC是统计正则化模型中调整正则化参数的有用标准，但它不能直接应用于核机器的参数调整，因为核函数仅定义特征空间中的内积。我们推导出一个特征值方程来计算KRIC和解决问题。通过使用Nyström近似，KRIC的参数调整的计算成本大大降低。由KRIC调整正则化参数的SVM或KLR的测试错误率与交叉验证或证据评估的测试错误率相当。KRIC的计算成本明显低于其他标准之一。
We present kernel regularization information criterion (KRIC), which is a new criterion for tuning regularization parameters in kernel logistic regression (KLR) and support vector machines (SVMs). The main idea of the KRIC is based on the regularization information criterion (RIC). Although the RIC is a useful criterion for tuning regularization parameters in statistical regularization models, it cannot be directly applied to parameter tuning for the kernel machines because kernel functions define only inner products in feature spaces. We derive an eigenvalue equation to calculate the KRIC and solve the problem. The computational cost for parameter tuning by the KRIC is reduced drastically by using the Nyström approximation. The test error rate of SVMs or KLR with the regularization parameter tuned by the KRIC is comparable with the one by the cross validation or evaluation of the evidence. The computational cost of the KRIC is significantly lower than the one of the other criteria.