A LSTM-based method for simulation execution validity evaluation

  • YuLun Wang,
  • Wei Li,
  • Jing Li 
  • Control and Simulation Center, Harbin Institute of Technology, Harbin 150080, China
  • Science and Technology on Complex System Control and Intelligent Agent Cooperation Laboratory, Beijing 100074, China
Cite as
Wang Y., Li W.,Li J. (2021). A LSTM-based method for simulation execution validity evaluation. Proceedings of the 33rd European Modeling & Simulation Symposium (EMSS 2021), pp. 327-332. DOI: https://doi.org/10.46354/i3m.2021.emss.045

Abstract

This paper improves the concept of simulation execution validity and a methodology to assess the complex simulation system. As the complexity of the simulation system gradually rises, the magnitude of the simulation data obtained continues to increase, which makes it extremely difficult for experts to provide expert knowledge to evaluate the execution validity of the simulation system. More precisely, the essence of execution validity is to dig out the hidden relationships in the simulation time series data and complete the classification task whether it is valid. Considering that machine learning can better complete the two tasks of mining data features and classification, this paper adopts long short-term memory, a neural network used to process time series data, to evaluate the execution validity. Finally, an experiment is conducted on a simulation system, and the results show that the evaluation method based on LSTM can accurately evaluate the validity of the simulation system, and can greatly improve the efficiency of evaluation.

References

  1. Boer, P. , Kroese, D. P. , Mannor, S. , & Rubinstein, R. Y. . (2005). A tutorial on the cross-entropy method. Annals of Operations Research, 134(1), 19-67.
  2. Elman, J. L. . (1990). Finding structure in time. Cognitive Science, 14(2), 179-211.
  3. Kingma, D. , & Ba, J. . (2014). Adam: a method for stochastic optimization. Computer Science.
  4. Lv, Z. , H Ding, Wang, L. , & Zou, Q. . (2021). A convolutional neural network using dinucleotide one-hot encoder for identifying dna n6-methyladenine sites in the rice genome. Neurocomputing, 422(4), 214-221.
  5. Reynolds, M. S. . (2010). Dynamically reconfigurable antennas for rfid label encoders/readers. US.
  6. Sundermeyer, M. , R Schlüter, & Ney, H. . (2012). LSTM Neural Networks for Language Modeling. Interspeech.
  7. Wahlbeck, K. , Tuunainen, A. , Ahokas, A. , & Leucht, S. . (2001). Dropout rates in randomised antipsychotic drug trials. Psychopharmacology, 155(3), 230-233.