Leopoldo Sarra

  • Doctoral Student
  • Theory Division
  • Room A.2.448
  • Phone +49 9131 7133 460
  • Email

I am a final year doctoral student exploring the uncharted territories of artificial intelligence and physics. With a broad range of research interests, I focus on theoretical physics topics such as statistical physics of disordered and complex systems, quantum information and computation. Alongside, I also enjoy working on various machine learning areas including information theory, unsupervised representation learning, generative models, and Bayesian experimental design.

During my PhD, I have been investigating how machine learning techniques can be employed to aid scientific discovery. I have been exploring the idea of an "Artificial Scientist", which can automatically learn from observations, understand relevant concepts, build new physical models, and design new experiments similar to how human scientists do. I have been actively pursuing theoretical work in this direction to provide the ingredients required for such a future invention.

My research interests also extend beyond the field of physics. I have conducted projects on more general research in machine learning and deep learning architectures. These projects, currently in the publishing process, are in collaboration not only within the Theory Division but also with computer science departments and industry, during my internship at DeepMind.

Please feel free to get in touch via email or connect on LinkedIn!

Automatic extraction of collective variables of a physical system

One of the most useful concepts in the analysis of a physical system is the notion of collective coordinates. In many cases, ranging from statistical physics to hydrodynamics, the description of a system can be dramatically simplified by considering only a few collective variables like the centre of mass, an order parameter or a flow field (think for example to solitons and optical vortices in the context of optics). However, in new situations, it is not clear a-priori which low-dimensional function is best suited as a compact description of the high dimensional data that the scientist observes.

In the work “Renormalized Mutual Information for Artificial Scientific Discovery” we developed a technique to find these quantities. The starting point was the use of statistics and information theory to “extract” the relevant quantities just by looking at many observations of the system. We generalized the idea of mutual information, a solid concept in information theory, to work for this purpose, i.e. to quantify the amount of information between a random variable and a deterministic continuous function of it. In addition, by parametrizing these features and performing optimization of the conveyed amount of information on the system, we showed that it is also possible to automatically extract collective variables.

We are currently pushing these ideas forward and applying them to statistical physics model.

Renormalized Mutual Information for Artificial Scientific Discovery
Leopoldo Sarra, Andrea Aiello, and Florian Marquardt
Phys. Rev. Lett. 126, 200601 – (2021)

Bayesian Experimental Design for Quantum Many-body systems

Bayesian experimental design is a technique that allows to efficiently select measurements to characterize a physical system, by maximizing the expected information gain. Recent developments in deep neural networks and normalizing flows allow for a more efficient approximation of the posterior and thus the extension of this technique to complex high-dimensional situations. We are investigating how this approach holds promise for adaptive measurement strategies to characterize present-day quantum many-body platforms. For example, we focus on arrays of coupled cavities and qubit arrays. Both represent model systems of high relevance for modern applications, like quantum simulations and computing, and both have been realized in platforms where measurement and control can be exploited to characterize and counteract unavoidable disorder. 

Concept Extraction: Discovering quantum circuits components with program synthesis

Despite rapid progress in the field, it is still challenging to discover new ways to take advantage of quantum computation: all quantum algorithms need to be designed by hand, and quantum mechanics is notoriously counterintuitive. In this paper, we study how artificial intelligence, in the form of program synthesis, may help to overcome some of these difficulties, by showing how a computer can incrementally learn concepts relevant for quantum circuit synthesis with experience, and reuse them in unseen tasks. In particular, we focus on the decomposition of unitary matrices into quantum circuits, and we show how, starting from a set of elementary gates, we can automatically discover a library of new useful composite gates and use them to decompose more and more complicated unitaries.

Discovering quantum circuits components with program synthesis
Leopoldo Sarra, Kevin Ellis, Florian Marquardt

Exploration in Reinforcement Learning

Scientists are agents that performs experiments to improve their model of the world. Their reward is mainly due to the observation of some new phenomenon and the improvement of their scientific model. In Reinforcement Learning, a similar situation happens when the reward is sparse, and it is useful to design strategies to explore the environment independently. Curiosity-driven and novelty-based techniques allow to intrinsically motivate the agent to explore the environment, in a similar way as a human scientist would do.
In this paper, we developed a particular novelty-based techniques, suitable for challenging exploration tasks like 3D-environments and Atari video games.

Unlocking the Power of Representations in Long-term Novelty-based Exploration
Alaa Saade, Steven Kapturowski, Daniele Calandriello, Charles Blundell, Pablo Sprechmann, Leopoldo Sarra, Oliver Groth, Michal Valko, Bilal Piot

Less recent works:

Quantum Information

Experimental semi-device-independent tests of quantum channels

Publication: I. Agresti, G. Carvacho, L. Sarra, R. Chaves, F. Buscemi, M. Dall'Arno, F. Sciarrino


Spin glasses

Study of longitudinal fluctuations of the Sherrington–Kirkpatrick model

Publication: G. Parisi, L. Sarra, L. Talamanca

- For information about the work on "Renormalized Mutual Information" you can look at the tutorial notebook, where you can also find the complete implementation.

- You can find more information on the theme of Artificial Scientific Discovery by watching the videos of the talks of the workshop Prof. Florian Marquardt and I organized during Summer 2021.



- For more information about Machine Learning in Physics, please look at the online courses offered by the head of our group, Prof. Florian Marquardt, both the introductory course and the advanced one .

- For information about hbaromega, the local student chapter of Optica (formerly OSA) and EPS Young Minds section, organizing scientific outreach and networking events please visit its website here or write us an email at hbaromega@mpl.mpg.de.

MPL Newsletter

Stay up-to-date with MPL’s latest research via our Newsletter. 

Current issue: Newsletter No 28 - April 2023

Click here to view previous issues.

MPL Research Centers and Schools