Manuscript received May 30, 2023; revised June 20, 2023; accepted July 5 2023; published February 4, 2024
Abstract—In this work, we model a double pendulum system with deep neural networks based on a data set generated from video recordings. For comparison, a similar model is made by describing the system with differential equations. Actually compared are the capabilities of both models in predicting the next 2s of double pendulum motion using information about the previous second. In addition, both models are compared by their ability to make predictions in specific error margins. Results show that deep learning-based approaches give much better predictions, where the best deep learning-based model could predict the next 1.5s in a specified error margin, while the best differential equation-based one only 0.12s, all other metrics agree with this result as well.
Keywords—curriculum learning, deep learning, differential equations, double pendulum, LSTM, teacher forcing
Cite: Reinis Freibergs, Ēvalds Urtāns, Ansis Ēcis, and Henrik Gabrielyan, "LSTM Rollout Curriculum Using Double Pendulum," International Journal of Machine Learning vol. 14, no. 1, pp. 12-17, 2024.
Copyright © 2024 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).