I mentioned in a recent post that k-fold cross-validation could be parallelized in Matlab through parfor and that this strategy works quite well. I’ve updated my two GLM packages (boosted GAMs and GLMs with Gaussian priors) to take advantage of this strategy. On quad core and 6-core computers, you can expect improvements of around 2x to 3x in speed, which can easily translate into days when working with large datasets. They’re available on Matlab Central.