Noise-robust optimization of quantum machine learning models for polymer properties using a simulator and validated...

Area of Science:

• Materials Science
• Quantum Computing
• Machine Learning

Background:

• Predicting polymer material properties from molecular descriptors is crucial.
• Existing quantum machine learning methods struggle with parameter optimization due to stochastic variations in quantum circuit outputs.

Purpose of the Study:

• To develop quantum machine learning techniques for accurate polymer property prediction.
• To enable robust parameter optimization in the presence of quantum sampling variability.
• To demonstrate the feasibility of training quantum models on real quantum hardware.

Main Methods:

• Investigated quantum circuits (multi-scale entanglement renormalization ansatz) for improved prediction accuracy without parameter increase.
• Employed stochastic gradient descent with the parameter-shift rule for robust gradient calculation.
• Trained and evaluated the quantum machine learning model on an ion-trap quantum computer and simulator.

Main Results:

• The multi-scale entanglement renormalization ansatz enhanced prediction accuracy without additional parameters.
• Stochastic gradient descent with the parameter-shift rule proved robust to sampling variability.
• The quantum model achieved comparable performance on both the quantum computer and simulator.

Conclusions:

• The proposed quantum machine learning approach enables effective parameter optimization despite stochastic variations.
• The study validates the training of quantum circuits on actual quantum hardware with performance parity to simulators.
• This work advances the application of quantum machine learning in materials science.