Talk, Austin Peay State University, Machine Learning Colloquium, Clarksville, Tennessee
Gaussian Processes have been utilized for a number of regression problems that we cannot assume any distribution on. The nonparametric qualities of gaussian processes allow for the development of a bayesian prior utilizing the observed points in the development of a posterior distribution. This method of model generation can be extended to a classification problem by applying the gaussian prior to a logistic regression problem. In this talk, I discuss examples of Gaussian Process Regression as well as an application on the MNIST dataset. The method is compared against other popular ML methods such as SVM, KNN, and MLP.