Fig. 3.

Effect of Max features and hash size on number of nodes (per tree), memory usage and run time for the two random forest implementations with 1000 trees. Data points are average values of the four datasets, and although imperceptible due to minuscule values: error bars are pooled standard deviations