Fig. 2From: Large sample size and nonlinear sparse models outline epistatic effects in inflammatory bowel diseasePerformance using different random subsets containing 10%, 20%, 40%, 60%, 80%, and 100% of the dataset. Models shown are the best additive model (\(\text {L}_2\) penalty), NNbiosparse and NNlinear. Performances are measured using ten different threefold cross-validation runs, using identical splits for the three modelsBack to article page