Meeting Notes 1/24

[panalluri]

Super-res:

• edit loss function s.t. equal weightage is given to all scales (see Yue and Jingwen's paper)
• Try rescaling inputs to NN
• Try different types of models (e.g. U-nets, GANs)
• Weights and biases --> start documenting
• Explore outputting means and variances from NN, like paper you read (figure out the paper?)
• lit review: come with two papers a week to discuss in 1 on 1 meetings --> keep thinking about which is the best to reproduce

Diagnostics:

• Find units for x, y axes in power spectra (hint: parseval's theorem)
• Slice data before filtering? Are the edges sensitive to this?
• Plot air-sea fluxes for all these filter scales
• Main goal: to understand how small-scale phenomena affects air-sea fluxes --> answer open Qs regarding small-scale processes from paper (Section 4.1 of Air-Sea Interaction submesoscale: physics and impact from Lorentz Center)