PennyLane–Qiskit Plugin: A Protocol for Integrating Noisy, Fake, and Real Quantum Backends

Abstract

Quantum computing frameworks like PennyLane, Qiskit, and others offer powerful tools for developing quantum algorithms. However, integrating their devices can be challenging for newcomers, especially given the frequent updates in frameworks like Qiskit, which can create significant barriers to learning and thus slow down the growth of quantum computing and quantum machine learning. The PennyLane–Qiskit plugin is a particularly interesting framework that allows researchers and practitioners interested in quantum computing and quantum machine learning to implement and simulate realistic models in a simple and efficient way. This work provides a practical method to use the PennyLane–Qiskit plugin, focusing on three main device types: the Aer simulator with noise models, IBM fake backends (with FakeManilaV2 as a representative example), and real IBM Quantum hardware accessed remotely (with ibm_torino used in this study). We demonstrate how to set up and use each of these devices within PennyLane using a simple example of a quantum machine learning regression task on simulated stock price data. Our goal is to simplify the process for all interested users, especially (but not only) beginners by clearly explaining and demonstrating how to use each of these device options while highlighting best practices for effectively leveraging PennyLane and Qiskit together via Pennylane–Qiskit plugin. This work aims to provide a turnkey solution enabling users with only minimal foundational knowledge to quickly start running meaningful simulations. The three devices implementation will be explained with the code here. The accompanying full code will be provided to ensure that all results can be easily reproduced, including with users’ own datasets.