Deprecate and move ad_hoc dataset to another place

[adekusar-drl]

What is the expected enhancement?

While most of the datasets have been deprecated in #319 the ad_hoc dataset is available in the module. This dataset is used in a few places and this issue is created to consider deprecation of this dataset as well. Perhaps, it should be kept somewhere else.

The places where ad_hoc is used so far:

• The readme note in the repo.
• QuantumKernel tutorial.
• Quantum Machine Learning course: https://learn.qiskit.org/course/machine-learning/introduction
  Update: Mar 2024, the content of this item above has now been archived, so I do not think its relevant anymore

Before the dataset is deprecated the above mentioned places must be updated.

[andrenmelo]

FYI I'm planning to rewrite ad_hoc and also and generalize to several qubits for my own research. Happy to help port it somewhere else too.

[adekusar-drl]

While ad_hoc is still here, you can open PRs with updates. It will take some time to decide how to proceed since the dataset is in use.

[edoaltamura]

@adekusar-drl where are the ad-hoc-like datasets for qiskit-community usually stored? Is the overall idea to move datasets into a different repository and add that as a dependency of machine-learning for tests and tutorials?

[edoaltamura]

@adekusar-drl @woodsp-ibm #902 extends the ad_hoc dataset, but what would you advise regarding the deprecation aspect? I saw there used to be more datasets, then removed in #319. What was the reason for deprecating?
While iris or wine can be fetched from elsewhere in principle, ad_hoc is 'unique' to QML as it comes from the Havlicek+2019 paper, so it might be worth keeping in the repo.

[adekusar-drl]

Honestly, I don't remember the details of the original issue, but the datasets from scikit-learn are easy to deal with and that's why they have been removed. Yes, you are right, ad_hoc is a unique one and since then I have change my mind I'd keep it as a pure quantum dataset.

[edoaltamura]

Great! In that case, #902 closed this issue.