Jun 27, 2013 — Image Registration — When we talk about registration, we are talking about a transform of one image into another image’s space by way of transforms.  What does this mean exactly? If each image is a matrix of numbers, the “transform” is another matrix that... Read more

Jun 26, 2013 — Generalized Linear Models (GLM) — If you think of regression (e.g., linear with gaussian distributions) and classification (logisitic with bernoulli) as cousins, they belong under the larger family of General Linear Models.  A GLM is a broad family of models that can all be simplified to a... Read more

Jun 25, 2013 — Laplace Smoothing — Why do we need Laplace Smoothing? Let’s return to the problem of using Naive Bayes to classify a recipe as “ice cream”(1) or “sorbet” (2) based on the ingredients (x).  When we are calculating our posterior probabilities, or the probability... Read more

Jun 25, 2013 — Naive Bayes — Naive Bayes is a supervised, probabilistic classification method for discrete classes that makes some strong assumptions about the data, but works pretty well.  If you think back to regression,  regression, we were trying to predict p(y x), and this is... Read more