Quantic has got a new paper out!
arXiv:2106.14059, by QUANTIC members @quantumAPS and @j_i_latorre. In this work, we explore the implementation of the data re-uploading quantum classifier on an ion-trap quantum computer.
Ion-traps are hard to scale, but the control in small systems is very good. This suits the requirements of our single-qubit system. The repeated application of high-fidelity single-qubit gates allows us to classify classical data on a completely quantum framework
The training is done in two steps:
1) A classical simulation is performed to obtain some theoretical optimal parameters
2) A further experimental optimization is carried to mitigate experimental errors
The experimental optimization is a crucial step to improve the performance of the quantum classifier. Theoretical results (red star) are not the optimal experimental configurations (blue area)
As in previous data re-uploading works, we can see how increasing the number of layers improves the final results. Top: 1, 2, 3 data re-uploadings. Bottom: comparison for 4 re-uploadings in the QPU and simulation.
More datasets can be classified successfully, for example:
A comparison between experimental results against classical algorithms is possible. As you can see, the experimental result (QPU column) is comparable to NNs and SVCs. This column will be filled in the future.
This work proves that single-qubit classifiers can work also in the laboratory with actual quantum devices. The high accuracy of ion-trap devices allows accomplishing these results. Here you can see a schematic description of the experiment.

To the best of our knowledge, this is the first functional quantum classifier with single-qubit systems.
Thank you very much to the team from @quantumlah and researchers Tarun Dutta, Manas Mukherjee, Jasper Phua Sing Cheng.