Automating Adjoints with Algorithmic Differentiation
HTML-код
- Опубликовано: 26 авг 2024
- A derivation of how to use implicit differentiation in reverse-mode AD (forward-mode is similar) to compute adjoints for a generic solver. Shows how to use vector Jacobian products for efficient calculation. Ends with a simple example of how you could implement in Julia (or in another language).
Very interesting video. It is like the Chain Rule, attempting to find the derivative in a composition of functions. Thank you so much !!!!!!
Thank you! Waiting to another awsome video :)