基于局部深度核学习的非线性支持向量机的有效预测
时间:2018-08-16 10:51 来源:毕业论文 作者:毕业论文 点击:次
| 摘要本文旨在加速非线性支持向量机的预测，并且能够在可接受的范围内保证分类的准确性。我们已经概括出局部多核学习，以便于起学会基于树状的高文、稀疏的原始特征嵌入。 因为非线性支持向量机的预测成本会随着训练集规模的增大而线性的增长，所以这对于非线性支持向量机来说呈现出巨大的挑战。因此，即使非线性支持向量机已经在多基准任务上定义出最先进的技术，但它们在现实世界的用途仍然有限。在可接受的分类精确度的情形之下，为非线性支持向量机的预测开发了局部深度核学习方法。最初的分类从支持向量的数量和训练集的规模上减少了预测的成本。在基准数据集的实验表明，在减少预测代价方面，局部深度核学习可以比RBF-SVMs在某些情况下的效果高出三个数量级。此外，局部深度核学习对于加速非线性支持向量机的预测可以完成更好的分类准确性。局部深度核学习也比其它一些近似核的技术要显著的更加优秀，比如随机傅里叶特征等，因为它更注重决策的范围，而不是在空间内随机的建立内核。局部深度核学习也比缩减集法更加快速，因为它的树状结构特征可以在对数级上计算。27077 关键词 深度学习 局部多核学习 局部深度核学习 优化局部深度核学习 毕业论文设计说明书外文摘要 Title Efficient prediction of non-linear SVM based on local deep kernel learning Abstract This paper aims to speed up non-linear SVM prediction while maintaining classification accuracy above an acceptable limit. We generalize Localized Multiple Kernel Learning so as to learn a tree-based primal feature embedding which is high dimensional and sparse. This presents a significant challenge for non-linear SVMs since their cost of prediction can grow linearly with the size of the training set. Thus, even though non-linear SVMs has defined the state-of-the-art on multiple benchmark tasks, their use in real world applications remains limited. We develop a Local Deep Kernel Learning (LDKL) technique for efficient non-linear SVM prediction while maintaining classification accuracy above an acceptable threshold. Primal based classification decouples prediction costs from the number of support vectors and the size of the training set。Experiments on benchmark data sets reveal that LDKL can reduce prediction costs by more than three orders of magnitude in some cases with a moderate sacrifice in classification accuracy as compared to RBF-SVMs. Furthermore, LDKL can achieve better classification accuracies for speeding up non-linear SVM prediction. LDKL is significantly better than kernel approximation techniques, such as Random Fourier Features , as it focuses on the decision boundary rather than modeling the kernel everywhere in space. LDKL can also be much faster than reduced set methods as its tree-structured features can be computed in logarithmic time. Keywords Deep Learning Localized Multiple Kernel Learning Local Deep Kernel Learning 目 录 1 引言 1 2 深度学习 3 2.1深度学习的介绍 3 2.2深度学习的基本思想 3 2.3深度学习的目标与体系 4 2.4深度学习的训练过程 5 3 深度核学习 7 3.1 相关工作 7 (责任编辑：qin) |