Mitigating barren plateaus with transfer-learning-inspired parameter initializations

Variational quantum algorithms (VQAs) are widely applied in the noisy intermediate-scale era and expected to demonstrate advantage. However, training VQAs faces difficulties, one of which is so-called barren plateaus (BP) phenomenon, where gradients cost functions vanish exponentially with number qubits. In this paper, inspired by transfer learning, knowledge pre-solved tasks could be further used a different but related work efficiency improved, we report parameter initialization method mitigate BP. method, small-sized task solved VQA. Then ansatz its optimum parameters transferred larger sizes. Numerical simulations show that BP improve efficiency. A brief discussion on how can well also provided. This provides reference for mitigating BP, therefore, more practical problems.