We prove all randomized sampling methods produce outliers. Given a computable measure P over natural numbers or infinite binary sequences, there is no method that can produce an arbitrarily large sample such that all its members are typical of P. The second part of this dissertation describes a computationally inexpensive method to approximate Hilbertian distances. This method combines the semi-least squares inverse techinque with the canonical modern machine learning technique known as the kernel trick. In the task of distance approximation, our method was shown to be comparable in performance to a solution employing the Nystrom method. Using the kernel semi-least squares method, we developed and incorporated the Kernel-Subset-Tracker into the Camera Mouse, a video-based mouse replacement software for people with movement disabilities. The Kernel-Subset-Tracker is an exemplar-based method that uses a training set of representative images to produce online templates for positional tracking. Our experiments with test subjects show that augmenting the Camera Mouse with the Kernel-Subset-Tracker improves communication bandwidth statistically significantly.