Extreme learning machine (ELMs) and gradient-based multilayer perceptrons (MLPs) have competing performance, in terms of learning speed and model accuracy. Thus, with reference to a specific application, it is of interest to establish whether the performance of ELMs are better than those achieved with MLPs. To this end this paper proposes a unified framework for the investigation of ELMs and MLPs performance in real data sets. The tool is based on Google’s TensorFlow architecture and is able to overcome the lack of previously available ELMs libraries. A wide experimentation on both toy and benchmark datasets shows the usefulness of the tool.
tfelm: a TensorFlow Toolbox for the Investigation of ELMs and MLPs Performance / Andrea, Castellani; Cornell, S.; Falaschetti, L.; Turchetti, C.. - (2018), pp. 3-8. (Intervento presentato al convegno 2018 International Conference on Artificial Intelligence (ICAI'18) tenutosi a Las Vegas, Nevada, USA nel July 30 - August 02).
tfelm: a TensorFlow Toolbox for the Investigation of ELMs and MLPs Performance
L. Falaschetti;C. Turchetti
2018-01-01
Abstract
Extreme learning machine (ELMs) and gradient-based multilayer perceptrons (MLPs) have competing performance, in terms of learning speed and model accuracy. Thus, with reference to a specific application, it is of interest to establish whether the performance of ELMs are better than those achieved with MLPs. To this end this paper proposes a unified framework for the investigation of ELMs and MLPs performance in real data sets. The tool is based on Google’s TensorFlow architecture and is able to overcome the lack of previously available ELMs libraries. A wide experimentation on both toy and benchmark datasets shows the usefulness of the tool.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
