International Journal of Networking and Computing
Online ISSN : 2185-2847
Print ISSN : 2185-2839
ISSN-L : 2185-2839
Special Issue on the Eighth International Symposium on Computing and Networking
ExtraFerns: Fully Parallel Ensemble Learning Technique with Random Projection and Non-Greedy yet Minimal Memory Access Training
Shungo KumazawaKazushi KawamuraThiem Van ChuMasato MotomuraJaehoon Yu
Author information
JOURNAL OPEN ACCESS

2021 Volume 11 Issue 2 Pages 215-230

Details
Abstract
Training machine learning models on edge devices is always a conflict with power consumption and computing cost. This paper introduces a hardware-oriented training method called ExtraFerns for a unique subset of decision tree ensembles, which significantly decreases memory access and optimizes each tree in parallel. ExtraFerns benefits from the advantages of both extraTrees and randomFerns. As extraTrees does, it generates nodes by randomly selecting attributes and generating thresholds. Then, as randomFerns does, it builds ferns, which are decision trees that share identical nodes at each depth. In contrast to other ensemble methods using greedy optimization, ExtraFerns attempts global optimization of each fern. Experimental results show that ExtraFerns requires only 4.3% and 4.1% memory access for training models with 3.0% and 1.2% accuracy drops compared with randomForest and extraTrees, respectively. This paper also proposes applying lightweight random projection to ExtraFerns as a preprocessing step, which achieved a further accuracy improvement of up to 2.0% for image datasets.
Content from these authors
© 2021 International Journal of Networking and Computing
Previous article Next article
feedback
Top