- Good and robust adaptive step sizing for implicit methods!
- Handling non-convergence of Newton's method(s) with implicit methods. The Newton solver can return a code indicating success/failure. Code was modified to do that in other repositories. If failure, do what though? Take a smaller step?
- Initial guess of k values for nonlinear solves in implicit methods.
- L-stable Rosenbrock method.
- Look into extra steps taken when snapping in an adaptive solves.
- Tridiagonal Jacobian support? The matrix routine is already in
linalg.ccbut there isn't a way to access it with implicit methods.