Multiple Kernel Learning for Brain-Computer Interfacing
Wojciech Samek, Alexander Binder, Klaus-Robert Müller
备注:
Corrected manuscript
发表信息:
W. Samek, A. Binder, K.-R. Müller. Multiple Kernel Learning for Brain-Computer Interfacing. Proceedings of 35th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 7048-51, 2013
Combining information from different sources is a common way to improve classification accuracy in Brain-Computer Interfacing (BCI). For instance, in small sample settings it is useful to integrate data from other subjects or sessions in order to improve the estimation quality of the spatial filters or the classifier. Since data from different subjects may show large variability, it is crucial to weight the contributions according to importance. Many multi-subject learning algorithms determine the optimal weighting in a separate step by using heuristics, however, without ensuring that the selected weights are optimal with respect to classification. In this work we apply Multiple Kernel Learning (MKL) to this problem. MKL has been widely used for feature fusion in computer vision and allows to simultaneously learn the classifier and the optimal weighting. We compare the MKL method to two baseline approaches and investigate the reasons for performance improvement.
Multi-task multiple kernel machines for personalized pain recognition from functional near-infrared spectroscopy brain signals
Harvard-MIT Health Sciences and Technology, Massachusetts Institute of Technology(麻省理工学院哈佛-麻省理工健康科学与技术项目);MIT Media Lab, Massachusetts Institute of Technology(麻省理工学院媒体实验室);Harvard Medical School(哈佛医学院);Boston Children’s Hospital(波士顿儿童医院);Massachusetts General Hospital(麻省总医院)
Brain-Network Clustering via Kernel-ARMA Modeling and the Grassmannian
University at Buffalo, The State University of New York (SUNY)(纽约州立大学布法罗分校);Drexel University(德雷塞尔大学);Perelman School of Medicine, University of Pennsylvania(宾夕法尼亚大学佩雷尔曼医学院)
Network Clustering Via Kernel-ARMA Modeling and the Grassmannian The Brain-Network Case
University at Buffalo, The State University of New York (SUNY)(纽约州立大学布法罗分校);Drexel University(德雷塞尔大学);Perelman School of Medicine, University of Pennsylvania(宾夕法尼亚大学佩雷尔曼医学院)
Northeastern University(东北大学);University College London(伦敦大学学院)
目前,神经科学界对大脑执行反向传播(BP)的可能性存在争议。为了更好地模拟大脑,已经提出了一种仅使用“单次向前传递”来训练网络$\textit{one layer at a time}$的方法,作为绕过BP的替代方案;我们将这些网络称为“逐层”网络。我们通过回答两个悬而未决的问题来继续分层网络的工作。首先,$\textit{do they have a closed-form solution?}$其次,$\textit{how do we know when to stop adding more layers?}$这项工作证明了内核均值嵌入是实现网络全局最优的封闭形式权重,同时驱动这些网络收敛到一个高度理想的内核进行分类;我们称之为$\textit{Neural Indicator Kernel}$。
目前,神经科学界对大脑执行反向传播(BP)的可能性存在争议。为了更好地模拟大脑,已经提出了一种仅使用“单次向前传递”来训练网络\textit{one layer at a time}的方法,作为绕过BP的替代方案;我们将这些网络称为“逐层”网络。我们通过回答两个悬而未决的问题来继续分层网络的工作。第一,$\textit{do they have a closed-form solution?}$第二,$\textit{how do we know when to stop adding more layers?}$这项工作证明了核均值嵌入是实现网络全局最优的封闭形式权重,同时驱动这些网络收敛到一个高度理想的分类核;我们称之为$\textit{Neural Indicator Kernel}$。