Machine learning (ML) is the study of algorithms and statistical models which can extract information hidden in a dataset and make predictions on unseen data from the same underlying probability distribution. In recent years ML techniques - and in particular deep neural networks - have made an enormous progress and have today a wide range of applications both in industry and in research.

In the range of activities of IN2P3, ML techniques are increasingly used in High Energy Physics, for data processing, as well as in Nuclear Physics, e.g. to solve nuclear many-body problems in complex energy landscapes.

Quantum machine learning (QML) is the intersection between machine learning and quantum computing and it has attracted considerable attention in recent years. The idea is that, since quantum computing is expected to bring us exponential speedups, usual machine learning tasks might as well be improved when executed on quantum computers.

Since the original algorithm by Harrow-Hassidim-Lloyd (HHL), a variety of ML algorithms for quantum computers have been proposed. These HHL-based “full” quantum ML techniques require high-depth quantum circuits which, in turn, would require large noiseless quantum hardware that is not available today.

In the present NISQ era, a realistic approach is to use hybrid techniques. In classical-quantum hybrid algorithms optimization is performed classically while the ML model is implemented by low-depth parametric quantum circuits. Analytic differentiation techniques are also available for such parametric circuits, allowing for gradient descent optimization.

The main goal of this project is to explore the future applications of QML in different areas of IN2P3 activity. More specifically, proof-of-principle applications will be provided, ranging from nuclear many-body problems to event recognition in multi-particle detectors.