Evaluating Teacher Forcing and Curriculum Learning in Recurrent Models for Oceanographic Time Series
Resumo
Sea level prediction is vital for port and coastal operations, where short-term forecasting accuracy is critical for navigation and planning. Autoregressive neural models are powerful tools for this task, especially when combined with training strategies like Teacher Forcing and Curriculum Learning. This study analyzes their impact on RNN performance for short-term sea level forecasting using high-frequency data from the Port of Santos, Brazil. We compare standard autoregressive training, Teacher Forcing, and Curriculum Learning across MAE, RMSE, and R² metrics, focusing on long prediction windows. Results show that low levels of teacher forcing improve convergence and reduce error over long horizons, highlighting its value for precision and long-term stability.Referências
Accarino, G., Chiarelli, M., Fiore, S., Federico, I., Causio, S., Coppini, G. and Aloisio, G. (2021) “A multi-model architecture based on Long Short-Term Memory neural networks for multi-step sea level forecasting,” Future Generation Computer Systems, vol. 124, pp. 1–9, November. [link]
Goodfellow, I., Bengio, Y. and Courville, A. (2016) Deep Learning, MIT Press, USA.
Goyal, A., Lamb, A., Zhang, Y., Zhang, S., Courville, A. and Bengio, Y. (2016) “Professor forcing: A new algorithm for training recurrent networks,” In: Advances in Neural Information Processing Systems, vol. NIPS, pp. 4608–4616.
Kartal, E. and Altunkaynak, A. (2024) “Empirical-singular-wavelet based machine learning models for sea level forecasting in the Bosphorus Strait: A performance analysis,” Ocean Modelling, vol. 188, p. 102324, April. [link]
Li, B., Yin, J., Zhang, A. and Zhang, Z. (2018) “A precise tidal level prediction method using improved Extreme Learning Machine with sliding data window,” In: 2018 37th Chinese Control Conference (CCC), IEEE, pp. 1787–1792.
Marengo, J. A., Muller-Karger, F., Pelling, M. and Reynolds, C. J. (2019) “The METROPOLE Project – An Integrated Framework to Analyse Local Decision Making and Adaptive Capacity to Large-Scale Environmental Change: Decision Making and Adaptation to Sea Level Rise in Santos, Brazil,” In: Climate Change in Santos Brazil: Projections, Impacts and Adaptation Options, Cham: Springer International Publishing, pp. 3–15.
Pang, T. Y., Ding, B., Liu, L. and Sergiienko, N. (2023) “Short-Term Sea Surface Elevation Prediction Using Deep Learning Methods,” In: Volume 5: Ocean Engineering, American Society of Mechanical Engineers.
Tur, R., Tas, E., Haghighi, A. T. and Mehr, A. D. (2021) “Sea level prediction using machine learning,” Water (Switzerland), vol. 13, no. 24.
Vicens-Miquel, M., Tissot, P. E. and Medrano, F. A. (2024) “Exploring Deep Learning Methods for Short-Term Tide Gauge Water Level Predictions,” Water, vol. 16, no. 20, p. 2886, October 11. [link]
Williams, R. J. and Zipser, D. (1989) “A Learning Algorithm for Continually Running Fully Recurrent Neural Networks,” Neural Computation, vol. 1, no. 2, pp. 270–280.
Goodfellow, I., Bengio, Y. and Courville, A. (2016) Deep Learning, MIT Press, USA.
Goyal, A., Lamb, A., Zhang, Y., Zhang, S., Courville, A. and Bengio, Y. (2016) “Professor forcing: A new algorithm for training recurrent networks,” In: Advances in Neural Information Processing Systems, vol. NIPS, pp. 4608–4616.
Kartal, E. and Altunkaynak, A. (2024) “Empirical-singular-wavelet based machine learning models for sea level forecasting in the Bosphorus Strait: A performance analysis,” Ocean Modelling, vol. 188, p. 102324, April. [link]
Li, B., Yin, J., Zhang, A. and Zhang, Z. (2018) “A precise tidal level prediction method using improved Extreme Learning Machine with sliding data window,” In: 2018 37th Chinese Control Conference (CCC), IEEE, pp. 1787–1792.
Marengo, J. A., Muller-Karger, F., Pelling, M. and Reynolds, C. J. (2019) “The METROPOLE Project – An Integrated Framework to Analyse Local Decision Making and Adaptive Capacity to Large-Scale Environmental Change: Decision Making and Adaptation to Sea Level Rise in Santos, Brazil,” In: Climate Change in Santos Brazil: Projections, Impacts and Adaptation Options, Cham: Springer International Publishing, pp. 3–15.
Pang, T. Y., Ding, B., Liu, L. and Sergiienko, N. (2023) “Short-Term Sea Surface Elevation Prediction Using Deep Learning Methods,” In: Volume 5: Ocean Engineering, American Society of Mechanical Engineers.
Tur, R., Tas, E., Haghighi, A. T. and Mehr, A. D. (2021) “Sea level prediction using machine learning,” Water (Switzerland), vol. 13, no. 24.
Vicens-Miquel, M., Tissot, P. E. and Medrano, F. A. (2024) “Exploring Deep Learning Methods for Short-Term Tide Gauge Water Level Predictions,” Water, vol. 16, no. 20, p. 2886, October 11. [link]
Williams, R. J. and Zipser, D. (1989) “A Learning Algorithm for Continually Running Fully Recurrent Neural Networks,” Neural Computation, vol. 1, no. 2, pp. 270–280.
Publicado
29/09/2025
Como Citar
MARUM, Tiago H.; AGRA, Ronney; BARROS, Marcel Rodrigues de; COSTA, Anna Helena Reali; COZMAN, Fábio Gagliardi; LOFRANO, Fábio Cunha; KUROKAWA, Fernando Akira.
Evaluating Teacher Forcing and Curriculum Learning in Recurrent Models for Oceanographic Time Series. In: ENCONTRO NACIONAL DE INTELIGÊNCIA ARTIFICIAL E COMPUTACIONAL (ENIAC), 22. , 2025, Fortaleza/CE.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2025
.
p. 986-996.
ISSN 2763-9061.
DOI: https://doi.org/10.5753/eniac.2025.14299.
