RDSF: Everything at Same Place All at Once - A Random Decision Single Forest

Random Forest is a widely-used machine learning approach. This work presents a novel graph representation called Random Decision Single Forest (RDSF) for Random Forests (RF). RDSF utilizes binary decision diagrams (BDD) to overcome challenges in RF implementations. It provides improved scalability, reduced execution time, and control over input data order compared to previous methods. The paper outlines the proposed mapping flow and experimental results, demonstrating the efficiency of RDSF for both numerical and categorical datasets. The RDSF significantly decreases generation time by up to two orders of magnitude and reduces inference time by one order of magnitude, as compared to the ADD-based approach.