MVPA_tutorial

MVPA tutorial - Rogers lab brain imaging unit [Wiki page]

I am organizing the brain imaging unit at the Knowledge and Concepts Lab, directed by Professor Tim Rogers. This is a tutorial (in progress) that introduces people to MVPA methods.

Neuroimaging data analysis is usually underdetermined. For example, a typical fMRI data might has 100,000 features (voxels) with only a few hunderds of training examples (stimuli presented). To tackle this issue of underdeterminacy while fitting the whole brain model (i.e. without pre-defining ROIs), we tend to use sparse methods, such as the Logistic LASSO, which will be the main focus of this tutorial.

The Brain Imaging Unit - Meeting

• You can access the meeting schedule from the wiki page.

• Where and when: Wednesday 4pm at Psych 634

Tutorial Content

• Compare L1 & L2 penalty: Linear Regression
• Logistic Regression
• Compare L1 & L2 penalty: Logistic Regression
• Iterative Lasso
• Group Lasso, SOS Lasso
• Readings

Dependencies

• Matlab
• Glmnet: A Matlab toolbox for fitting the elastic-net for linear regression, logistic and multinomial regression, Poisson regression and the Cox model.

Reference:

• Glmnet for Matlab (2013) Qian, J., Hastie, T., Friedman, J., Tibshirani, R. and Simon, N. http://www.stanford.edu/~hastie/glmnet_matlab/