@Nelimee In light of the challenges you’ve outlined regarding the use of variational quantum algorithms for solving PDEs, do you see any emerging technologies or methodologies in quantum computing that could potentially overcome these obstacles? Specifically, are there any advancements in quantum algorithm design or error correction techniques that might make variational quantum algorithms more reliable or efficient in the future?
This is a million-dollars (maybe billion?) question you are asking here. Disclaimer: the field is still evolving at a fast pace, and I only skimmed through a few articles on the subject in the last 2 years. I may be completely wrong in the following lines.
The only variational quantum algorithms (VQA) that may, in my opinion, eventually have an advantage are the ones with enough symmetries in the re-phrased optimisation problem to be free from Barren Plateau and to allow a costly-but-not-unreasonable exploration of the optimisation landscape with good hopes to find the global minimum.
Even in the above case, there are a lot of practical things that will hinder variational quantum algorithm performance.
For time-dependent PDEs, there is a fundamental limit in the approaches I know about that prevent any kind of exponential advantage. This will be the subject of another blog post, but this one will need more science that usual, so it will likely be posted last or close to the end of the series.
Error-correction will solve none of the issues I talk about in this series. Fault-tolerance will not either.
Quantum algorithm design might help, but the limits I talk about here are quite fundamental to the way VQA work. If you rephrase your problem without a cost function to minimise, (my understanding of the term says that) this is not a VQA anymore. The output of this core “re-phrasing as an opimisation problem” part is the root cause of the problem I write about in this blog post and Barren Plateau.
@Nelimee In light of the challenges you’ve outlined regarding the use of variational quantum algorithms for solving PDEs, do you see any emerging technologies or methodologies in quantum computing that could potentially overcome these obstacles? Specifically, are there any advancements in quantum algorithm design or error correction techniques that might make variational quantum algorithms more reliable or efficient in the future?
Cool Blog btw, I love the post titles!
This is a million-dollars (maybe billion?) question you are asking here. Disclaimer: the field is still evolving at a fast pace, and I only skimmed through a few articles on the subject in the last 2 years. I may be completely wrong in the following lines. The only variational quantum algorithms (VQA) that may, in my opinion, eventually have an advantage are the ones with enough symmetries in the re-phrased optimisation problem to be free from Barren Plateau and to allow a costly-but-not-unreasonable exploration of the optimisation landscape with good hopes to find the global minimum. Even in the above case, there are a lot of practical things that will hinder variational quantum algorithm performance. For time-dependent PDEs, there is a fundamental limit in the approaches I know about that prevent any kind of exponential advantage. This will be the subject of another blog post, but this one will need more science that usual, so it will likely be posted last or close to the end of the series. Error-correction will solve none of the issues I talk about in this series. Fault-tolerance will not either. Quantum algorithm design might help, but the limits I talk about here are quite fundamental to the way VQA work. If you rephrase your problem without a cost function to minimise, (my understanding of the term says that) this is not a VQA anymore. The output of this core “re-phrasing as an opimisation problem” part is the root cause of the problem I write about in this blog post and Barren Plateau.