Multi-GPU approach for big data mining
Krzysztof Jurczuk, Marcin Czajkowski, Marek Krętowski · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2019
This paper identifies scalability bounds of the evolutionary induced decision trees (DT)s. In order to conquer the barriers concerning the large-scale data we propose a novel multi-GPU approach. It incorporates the knowledge of the global DT induction and EA parallelization. The search for a tree structure and tests is performed sequentially by a CPU, while the fitness calculations are delegated to GPUs, thus the core evolution is unchanged. The results show that the evolutionary induction is accelerated several thousand times by using up to 4 GPUs on datasets with up to 1 billion objects.