I have been investigating uses for quantum machine learning, and have made a few working examples (variations of variational quantum classifiers using PennyLane). However, my issue now is its relationship with classical machine learning. At the moment (in my tests, at least), QML seems to not provide any major improvement in performance (compared to a classical network) and is significantly slower when running on real hardware.

I understand that this is a young field people are still exploring, but I'm curious as to why you would not just always use a classical ML algorithm for problems. Therefore, my questions are:

  • What benefits (or predicted benefits) are there using quantum machine learning?
  • Is there little benefit now, but the potential for performance increases when hardware improves?
  • It wouldn't surprise me to learn there are examples where QML outperforms classical ML. Here, my question is why is this the case? How would moving to a quantum regime improve performance?
  • 3
    $\begingroup$ QLM (as you are using the term here) is to machine learning as quantum algorithms are to classical algorithms. I feel like you could seamlessly remove all references to ML from this question, asking only about the relation between quantum algorithms and classical algorithms, and the answers would be essentially the same $\endgroup$
    – glS
    Aug 17 '20 at 15:53
  • 1
    $\begingroup$ Seconded gIS, also from what I've heard QML is still an extremely nascent field with not a lot of landmark results. I think HHL (which is not even QML really) is probably one of the biggest results. We don't really have the hardware yet to do larger tests + I believe one of the difficulties was loading data into the systems (which is only avoided in very specific use cases) $\endgroup$
    – C. Kang
    Aug 17 '20 at 16:07
  • $\begingroup$ This answer to another post may be answering the same question $\endgroup$ Mar 3 at 21:05

Check out these resources. It shows how QML might turn out to be in the future, see IBM Q for AI.

In case of quantum-inspired algorithms, when the dataset meets certain conditions, it might be better than the classical approaches, see Quantum-inspired algorithms in practice


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.