May I know if pennylane has support for second order optimizers (Gauss-Newton, Newton’s methods, Levenberg-Marquardt, etc.)? If not, would this be a feature that the community would be interested in?
Hi @syedfarhan, we do not have the optimizers that you mention. All of the optimizers we support are here.
It would be interesting to have at least one of these optimizers implemented. If you want to try implementing one of them it would be great!
Thanks for letting me know about it @CatalinaAlbornoz! I am using a second order optimizer in my code (but haven’t been able to get good results). I will try implementing one with pennylane when my code is stable.
Feel free to open a pull request on PennyLane when your code is stable @syedfarhan!
We also recently got a contributor who added an SPSA optimizer so I’m sharing it with you in case it’s helpful! Notice that this optimizer is only available on Master, and will be available on the next stable PennyLane release.
You can also use pytorch optimizer such as LBFGS wich has interface with Pennylane.
That’s true! If you’re using the PyTorch interface then you can use PyTorch optimizers.