Machine Learning for Neuroimagers
Overview
Machine Learning is a method of using data to train a classifier; this is called training data. The classifier is then provided with new data (also known as testing data), and it attempts to distinguish between different classes within the data based on the training data. The classifier’s performance is judged by its accuracy - how many of the testing data points it managed to correctly classify.
After working through an example with AFNI to learn the basics, we will begin to use a Matlab package called The Decoding Toolbox. This will expand our options of different classifiers to use, such as searchlight algorithms and representational similarity analysis (RSA). To begin reading an overview of machine learning and how it is used with fMRI data, click the Next
button; otherwise, select any of the chapters below to begin using either 3dsvm or The Decoding Toolbox.
Note
Thanks to Martin Hebart for helpful comments, especially about the statistical analysis of MVPA data. A useful series of PDFs and lectures can also be found here.
- Machine Learning: Introduction to Basic Terms and Concepts
- Machine Learning Tutorial #1: Basic Example with Support Vector Machines
- Machine Learning Tutorial #2: The Haxby Dataset
- Machine Learning Tutorial #3: Preprocessing
- Machine Learning Tutorial #4: Creating the Timing Files
- Machine Learning Tutorial #5: MVPA Analysis with The Decoding Toolbox
- Machine Learning Tutorial #6: Scripting
- Machine Learning Tutorial #7: Group Analysis
- Machine Learning Tutorial #8: Non-Parametric Analysis
- Machine Learning Tutorial #9: Representational Similarity Analysis
- Machine Learning Tutorial #10: Hyperalignment