This paper presents a study about the use of two deep neural networks (MLP and LSTM) in the Dynamic Helper System (DHS), which is a system that only exists in the EFC railway. The DHS assists a fully loaded train in sections with slopes and climbs. Therefore, this work shows that it is possible to predict the next action of the DHS using a classification algorithm, specifically, deep neural networks. The training and testing dataset is a real one, and the training task was performed using PyCUDA, Keras, and TensorFlow in a General Purpose GPU. Results using a real-world dataset indicate that deep neural networks can reach an accuracy above 99% and precision about 81% on predicting the next action of the DHS.
CITATION STYLE
Costa, J. P. A., Cortes, O. A. C., de Oliveira, V. B. S. N., & da Silva, A. C. F. (2019). Deep Neural Networks Applied to the Dynamic Helper System in a GPGPU. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 11508 LNAI, pp. 29–38). Springer Verlag. https://doi.org/10.1007/978-3-030-20912-4_3
Mendeley helps you to discover research relevant for your work.