Text

  • Study location R3-121 (Västerås)
Date
  • 2025-01-22 14:30–15:30

Leando Farina. Navigating Sparsity and Nonlinearity: Deep Learning for
Temporal Interpolation of Ocean Wave Spectra

Time: 2025-01-22, 14:30-15:30

Location: R3-121

Video link: https://mdu-se.zoom.us/j/69339369403 External link.

Participating: Leandro Farina (Federal University of Rio Grande do Sul (UFRGS), Porto Alegre, Brazil)

Abstract: The directional spectrum of ocean waves is a sparse function characterized by variable peaks and nonlinear evolution. Furthermore, observational data is often noisy, particularly in the higher-order moments of the spectrum. This talk explores the mathematical challenges associated with the temporal interpolation of these wave spectra. We present a comparative analysis of various interpolation techniques, ranging from
classical linear methods and radial basis functions to state-of-the-art deep neural networks. Using data acquired from oceanographic buoys in the Atlantic and Pacific Oceans, we apply these interpolation methods, analyzing the trade-offs between computational cost, accuracy, and the ability to capture the underlying nonlinear dynamics. We will discuss methods for quantifying the degree of nonlinearity present in the
spectra by analyzing residuals from linear trend fitting across different frequencies and directions. Finally we evaluate the effectiveness of each approach in modeling and predicting the temporal evolution of oceanwave behavior, highlighting the strengths and limitations of data-driven techniques in handling the inherent sparsity and nonlinearity of this problem.

More information

For more information about the research milieu, please contact: