First Learn tO Walk (FLOW)
Karina Ramos Musalem, Noah Rosenberg
Karina Ramos Musalem, Noah Rosenberg


Using a simple climate model and its algorithmic differentiation-generated adjoint, can we perform a state estimation and recover realistic control parameters?
Explorations
- Implementation of various gradient descent algorithms.
- How does the estimate change as we add and normalize controls?
Products
- Two tutorial notebooks with self-contained state estimation procedures:
- State Estimation in the Budyko-Sellers energy balance model using algorithmic differentiation in JAX
- State Estimation in the Budyko-Sellers energy balance model: Exploring the role of controls