RFAcc: a 3D ReRAM associative array based random forest accelerator

Random forest (RF) is a widely adopted machine learning method for solving classification and regression problems. Training a random forest demands a large number of relational comparison and data movement operations, which take long time when using modern CPUs. Accelerating random forest training using either GPUs or FPGAs achieves only modest speedups. In this paper, we propose RFAcc, a ReRAM based accelerator, to speed up random forest training process. We first devise a 3D ReRAM based relational comparison engine, referred to as 3D-VRComp, to enable parallel in-memory value comparison. We then exploit 3D-VRComp to construct RFAcc to speedup random forest training. Finally, we propose three optimizations, i.e., unary encoding, pipeline design, and parallel tree node training, to fully utilize the accelerator resources for maximized throughput improvement. Our experimental results show that, on average, RFAcc achieves 8564 and 16850 times speedup and 6.6 × 104 and 2.6 × 105 times energy saving over the training on a 4.2GHz Intel Core i7 CPU and a NVIDIA GTX1080 GPU, respectively.