Work Package II

Last updated on Jul 7, 2022

Inversion and prediction by neural networks

In this WP, novel DNN models adapted to the specificity of SAR displacement data (i.e. spatial and time varying data), will be proposed for physical parameters inversion and prediction problems. We will rely on generative networks, i.e. GAN, and recurrent networks such as LSTM or causal convolutional networks.

  • Task I: Physical parameters inversion & prediction by supervised neural networks learning

  • Task II: Semi-supervised modelling of physically relevant latent features

  • Task III: neural networks model interpretability and explainability

Yajing Yan
Yajing Yan
Associate Professor of Remote Sensing for Earth Observation

My research interests include Interferometry SAR, multi-temporal analysis, data inversion, data assimilation and machine learning.