Jo will give us an intro to SVM, and also this talk:

A new MVPA-er’s guide to fMRI datasets
fMRI datasets have properties which make the application of machine learning (pattern recognition) techniques challenging – and exciting! This talk will introduce some of the properties most relevant for MVPA, particularly the strong temporal and spatial dependencies inherent in BOLD imaging. These dependencies mean that some fMRI experimental designs are more suitable for MVPA than others, due, for example, to how the tasks are distributed within scanner runs. I will also introduce some of the necessary analysis choices, such as how to summarize the response in time (e.g., convolving with an HRF), which brain areas to include, and feature selection techniques.