1
$\begingroup$

Are there Quantum-enhanced Machine Learning algorithms that can be implemented via Qiskit in IBM Q Experience and obtain valuable inferences faster than their classical counterparts from datasets of let's say Kaggle?

Please also link important papers and articles that you think are worth reading. Thank you!

$\endgroup$
  • $\begingroup$ each post should contain a single question. Please edit your post to focus on a single issue, and ask the others on separate questions, as it is too broad as it stands. Also, you might find some information in this answer and this answer $\endgroup$ – glS Jun 3 at 19:24
  • $\begingroup$ if you are asking whether it is right now possible to use a quantum computer to solve practical problems faster than classical algorithms, no, that's not possible (at least not in any straightforward way that can be used by non-experts) $\endgroup$ – glS Jun 6 at 13:47
  • $\begingroup$ Is that because the data to be processed by quantum machines needs to be first encoded as amplitudes of quantum state which takes more time? $\endgroup$ – IDK IDK Jun 8 at 8:20
1
$\begingroup$

Quantum machine learning can help you to enhance classical machine learning algorithms by outsourcing difficult calculations to a quantum computer. You can also optimise quantum algorithms using classical machine learning architectures.

IBM researchers have developed a series of quantum algorithms that show how entanglement can improve AI classification accuracy. It is demonstrating a Quantum Classifier. It is available online on IBM Bluemix in the following link. Also IBM has demonstrated Hybrid quantum - classical neural networks with PyTorch and Qiskit online in the following documentation.

When you are working on a Quantum Classifier for a dataset, we need to first encode the data into the amplitudes of a quantum state. In fact, one needs to first normalise the data such that it can be represented as a vector on a high-dimensional Bloch sphere. Quantum routines to encode data in amplitudes, so called arbitrary state preparation routines, are known to do this with a runtime that is linear in the data size, and this is arguably the best our algorithm can do in terms of runtime, since the data is the input to the problem.

| improve this answer | |
$\endgroup$

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