Fig. 2From: LANDMark: an ensemble approach to the supervised selection of biomarkers in high-throughput sequencing dataA simplified overview of how LANDMark constructs and selects splitting rules at each node. Multiple models are evaluated at each node in each tree. Diversity between LANDMark trees is due to the random selection of training samples and features in each node, random initializations of most models, and the random selection of models which create partitions with equivalent information gainBack to article page