In many applications including health-care, finance and object recognition, data classification may be hindered by the existence of multiple contexts that produce an input sample. These contexts are generally hard to define, they are often interlaced and do not have sharp boundaries. Context-based classifiers offer the promise of increasing performance by allowing classifiers to become experts at classifying input samples of certain types. In this book, we introduce several models that can simultaneously learn the contexts as well as the classifiers for static, sequential and time-series data. We demonstrate the results on landmine detection from ground penetrating radar and electro-magnetic induction sensors, and show how choosing an appropriate context can simplify the classification problems.