Q-Trainer: A Software Package for Training Variational Quantum Algorithms with Quantum Error Mitigation

The training of variational quantum algorithms (VQA) on modern quantum computers inevitably involves large noises, and quantum error mitigation (QEM) can be used to mitigate such noises. However, currently, it requires much domain knowledge (of both VQA and QEM) to apply QEM to VQA in practice, making QEM less accessible to beginners in quantum computing. Motivated by this, we developed Q-Trainer, a Python package with high-level API for VQA training with QEM. Our package is built on i) PennyLane, a quantum software platform with many VQA implementations, and ii) Mitiq, a software library of QEM methods.

With Q-Trainer, users only need to define a VQA problem instance (e.g., a graph for QAOA-MaxCut), training configurations (e.g., device, optimizer, step size, number of epochs), and the preferred QEM method, then Q-Trainer can directly train a variational quantum circuit as requested. Notably, Q-Trainer is not only compatible with CPU/GPU backends, but also supports quantum computer backends through IBMQ and Amazon Braket services.