NeuralPDE.jl
NeuralPDE.jl copied to clipboard
Add a more performant approximation for directional derivatives
Directional derivatives are currently calculated by first calculating all the variable derivs and then performing a weighted sum.
This could be more efficient, especially with many dimensions.
Actions to take:
- [ ] Add a Nabla/Del operator to symbolics
- [ ] Recognise directional derivs and dispatch a more performant apporximation