Matlab implementation of the TCK
The time series cluster kernel (TCK) is a similarity measure and kernel for multivariate time series with missing data. The TCK is designed using an ensemble strategy to probabilistic models. SInce it is a kernel it can be used as input in many different learning algorithms, in particular in kernel methods such as SVMs and kPCA. Some strengths of the method are 1) no parameter-tuning and 2) robustness to missing data. Further details


Comments are closed.