Fast-Least-Squares

Fast cross-validation for least-squares methods and multi-class LDA

Link to arXiv.org preprint: http://arxiv.org/abs/1803.10016

Main functions

• fast_least_squares: implements the analytical cross-validation approach for least-squares models and multi-class LDA
• standard_lda_traintest: for comparative purposes, implements standard cross-validation for LDA and multi-class LDA

Simulations

The simulations reported in the paper can be reproduced using the scripts in the subfolder simulation. The function simulate_gaussian_data creates the multivariate Gaussian data. It is called from the scripts simulation1_binary_LDA and simulation2_multiclass_LDA which perform the cross-validation and permutations experiments.

MEEG data

The analyses performed on the Wakeman and Henson can be reproduced using the scripts in the subfolder WakemanHensonData. The data can be downloaded from the OpenfMRI website.

• WakemanHenson_preprocess: preprocesses the EEG/MEG data using Fieldtrip.
• WakemanHenson_run_permutations_binary_LDA: performs the permutations analysis for binary LDA
• WakemanHenson_run_permutations_multiclass_LDA: performs the permutations analysis for multi-class LDA
• results_WakemanHenson: collects the results of the permutation analyses and stores it in tables