Figure 2.

Timing experiments. Time (in seconds) for model fitting, over sub-samples of the celiac disease dataset, taken as the minimum time over 10 independent runs. The inset panel shows the results for 50,000 SNPs in more detail, note the different scales. For in-memory methods we included the time to read the data into memory. For SparSNP and glmnet we used a penalty grid of size 20, and a maximum model size of 2048 SNPs. LIBLINEAR (denoted “LL-L1”) and LIBLINEAR-CDBLOCK (denoted “LL-CD-L2”) induced one model with C = 1. LIBLINEAR-CDBLOCK used m = 50 blocks. For some datasets, glmnet and LIBLINEAR did not complete and these running times are not shown. HyperLasso is not shown since it took much longer to complete than the other methods

Abraham et al. BMC Bioinformatics 2012 13:88   doi:10.1186/1471-2105-13-88
Download authors' original image