Parallel tree-ensemble algorithms for GPUs using CUDA