Nonparametric methods in DMML usually refers to the use of finite, possibly small data sets to search large spaces of functions. Large means, in particular, that the elements of the space cannot be indexed by a finite-dimensional parameter. Thus, large spaces are typically infinite-dimensional–and then some.
KeywordsMean Square Error Kernel Regression Optimal Bandwidth Kernel Smoother Kernel Density Estimator
Unable to display preview. Download preview PDF.