Machine learning is a data-driven field, which needs massive computing resources. Quantum computation, on the other hand, can provide exponential speedups for some classical algorithms. Therefore, it is natural to combine the strengths of both fields to solve outstanding problems in industry and research. The project has three goals. The first goal is to use machine learning methods for the description of many-body quantum systems. In this part of the project, we will tackle some of the notable problems of many-body quantum mechanics with new tools that are emerging by adopting neural networks to quantum mechanical problems. The second goal is to use methods from many-body quantum mechanics to describe machine learning problems. We will address the problems of adversarial examples, uncertainty, and generalization from a new perspective, which is motivated by the success of tensor networks for a description of many-body quantum systems. The third and most ambitious goal is to combine the knowledge from quantum mechanics and machine learning to find novel applications of noisy intermediate-scale quantum devices with significant speedups for known classical algorithms. We will apply a combination of successful quantum-mechanical tools and advanced machine learning tools to find useful quantum algorithms that could demonstrate applied quantum advantage.

Research activity

Natural sciences and mathematics

Range on year

1,25 FTE (1,07 at UL FRI)

Research organisations

Faculty of computer and information science, UL

Faculty of mathematics and physics, UL

Researchers

Bojan Žunkovič (V)

Enej Ilievski (R)

Marko Robnik-Šikonja (R)

Marko Žnidarič (R)

Project phases and their realization

The project has four work packages. Work packages WP1 and WP2 explore the relation between quantum mechanics and machine learning in opposite directions and can run in parallel and also benefit from each other. They form the core of the project and will be prioritized. Part of the work package WP3 depends on the working pipeline of the package WP2. Therefore, we prioritize WP2 and start working on WP3 beginning of the third year of the project. The second reason is that we expect significant improvements to the NISQ devices in the next two years, which will increase the chances of performance improvements of quantum algorithms with respect to classical approaches. We plan the dissemination activities (WP4) after finished tasks and when we expect to have more time.

Project bibliographic references

Žunkovič, B., Heyl, M., Knap, M., & Silva, A. (2018). Dynamical quantum phase transitions in spin chains with long-range interactions: Merging different concepts of nonequilibrium criticality. Physical review letters, 120(13), 130601.

Lerose, A., Marino, J., Žunkovič, B., Gambassi, A., & Silva, A. (2018). Chaotic dynamical ferromagnetic phase induced by nonequilibrium quantum fluctuations. Physical review letters, 120(13), 130603.

Robnik-Šikonja, M. (2015). Data generators for learning systems based on RBF networks. IEEE transactions on neural networks and learning systems, 27(5), 926-938.

Kranjc, J., Orač, R., Podpečan, V., Lavrač, N., & Robnik-Šikonja, M. (2017). ClowdFlows: Online workflows for distributed big data mining. Future Generation Computer Systems, 68, 38-58.

Ljubotina, M., Žnidarič, M., & Prosen, T. (2017). Spin diffusion from an inhomogeneous quench in an integrable system. Nature communications, 8(1), 1-6.

Financed by

Slovenian Research Agency