Cerca nel sito

01. Quantum photonic technologies, artificial intelligence and complexity

Quantum photonic technologies, artificial intelligence and complexity

This activity is carried out in collaboration with Prof. Sciarrino (www.quantumlab.it) and Prof. Accounts of the Physics Department of Sapienza, “New Talents” of the CREF in the years 2005–2008. This line of research aims to strengthen collaboration with leading groups in the field of quantum technologies and applications related to artificial intelligence, as well as with regard to the involvement of students and the PhD course in the Department of Physics of Sapienza. The main objectives are the realization of new computational and quantum information experiments with photonic technologies, with strong support in related theoretical and computational studies.


Quantum technologies are undoubtedly one of the fields attracting the greatest strategic interest in the world. Witness the enormous investment by the United States, China, and the European community, with the involvement of large industrial companies such as IBM and Google.

There are numerous open problems related to quantum technologies, both from a fundamental and an applicative point of view. Despite a growing number of researchers and an increasingly active community, this research has not yet reached the stage of maturity to concretely impact society, or to radically change the paradigms of modern physical science. However, the impressive developments and convergence in quantum technologies of disciplines, such as artificial intelligence, photonics and complex systems, open horizons that cannot be overlooked.

In this context, CREF can play a strategic role at the national and international levels, thanks to the strong link with the Department of Physics of Sapienza where some of the leading scientists of the field are active (many of whom are former grantees of the institute with a promising community of students and young researchers.

CREF can be a reference center for the development of new concepts and new experiments, and also a nucleus of aggregation, thanks to its history and the facilities it makes available, for conferences and newly launched laboratories. The mix with other nearby realities, such as Sapienza University, CNR, and INFN, can create a center for highly innovative photonics with a high scientific impact due to coordination at Via Panisperna.

This line of research is framed in this context and begins an ambitious path to create a large-scale reality. The starting point is a new laboratory shared between CREF and the Department of Physics of Sapienza that envisages as its first objectives the creation of new computational machines at the frontier between classical and quantum systems to solve problems of social interest, such as those related to economics or to the optimization of complex systems, and to address the many fundamental aspects that quantum technologies open up.

Enrico Fermi was a pioneer in the creation and use of calculating machines, key examples of which are displayed in the CREF Museum. It is often argued that quantum systems can solve combinatorial optimization problems in a time that varies in a polynomial way with the size of the system. This possibility is often referred to as “Quantum Advantage” or even “Quantum Supremacy.” However, the practical implementation of these machines shows that a series of physical effects (such as the excitation of spurious energy states) makes the law of scale exponential, as in classical computers. The question then arises whether there is a real advantage in quantum computation applied to combinatorial problems. The solution to be pursued is the development of hybrid computational machines that use photonic quantum systems to accelerate the computation, but that provide the result of the computation in a robust classical form, which is not subject to decoherence and therefore can be immediately interfaced with traditional calculators. The first experimental evidence of this possibility have been reported by the participants in this project, who, thanks to photonics, have preliminarily demonstrated optical calculations with 105 spins [Pierangeli,2019], a scale never reached before. These “proof-of-concepts” will be developed extensively within the CREF leading to a new realization of photonic calculators. The fundamental physical problems related to the classical–quantum interface, the role of entanglement in many-body photonic systems, nonlinear effects and the collective modes of the systems that are intended to be used for the calculation will also be addressed.

The guideline in this context is the interdisciplinary use of notions from apparently distinct fields, such as artificial intelligence, photonics, physics of nonlinear systems and obviously quantum information. It is becoming increasingly clear that the use of machine learning is changing the way experiments are carried out in the laboratory. The complexity of experiments involving different interacting systems (electronic, optical, quantum) on various spatial and temporal scales, pushes scientists to an increasingly “data-driven” approach, which follows the trend observed in other fields such as economics and network theory. A practice increasingly in use is to enrich an experimental apparatus with a layer of artificial intelligence, aimed not only at the optimization of the observation parameters, but also at guiding the researcher in the most promising direction for observation and to report aspects that are not provided. The availability of novel computational resources and new paradigms leads to the design of experiments with a growing degree of innovation. Examples are large-scale experiments such as the “Big-Bell test,” [Abel2018] or innovative concepts, such as Ising’s machines [Pierangeli2019]. The development of experimental physical techniques in the field of photonics, supported by machine learning with paradigms such as “Reinforcement Learning”, will be one of the main objectives.

This line is part of a strongly international context mediated by initiatives such as the European Flagship for Quantum Technologies, or projects of the European Research Council and Future Emerging Technologies. The participants of this line develop their research in the context of collaborations involving major universities and research centers.

Another objective is the substantial involvement of students and the scientific community. The planned initiatives include workshops at the CREF, guided visits to the laboratories, internships, granting of degree and doctoral theses.

This line of research, intended to strengthen the scientific community on quantum technologies around the time of CREF Annual meetings will be organized at the CREF, involving the leading scientists of the field and numerous students. The purpose of the meetings will be to identify the fields of greatest interest, new challenges, and the development of new experiments and initiatives, such as research projects, with a view to  incubate new scientific ideas in which the new CREF is involved. Awards will also be created for promising young researchers and researchers.

The greatest challenge is the setting up of new laboratories at the CREF for newly conceived experiments in the field of quantum technologies and also in direct connection with the laboratories and students of the Physics Department Physics of Sapienza University degree course. Among the laboratories to be set up, we highlight, in particular:

– Quantum Causality and Technologies (QCT) laboratory: a compact and bright source of entangled states will be developed. This source will be used both for educational purposes related to the activity of the CREF Museum and for research activities. In particular, various quantum causality schemes will be created, and the QCT Laboratory will serve as a node within a network connected with the Department of Physics—Sapienza University of Rome.

– Laboratory of Classical and Quantum Optical Computing: computational machines with photonic technologies will be developed that also aimed at dissemination and research. In particular, various systems will be created with the paradigms described later and made available for the various applications, including for educational purposes.

The developments obtained by both laboratories will then be combined to carry out optical computation experiments using quantum states of light as a resource.

Hybrid quantum and classical computing systems are explored to develop and experimental test new computational paradigms applied to real world “data-driven” combinatorial problems. The construction and implementation “in the spirit of Fermi” of actual computers based on optical technologies is expected to be put online and made accessible to all for different applications.

Among the planned activities is the construction of Ising machines on a scale never achieved before. Ising machines solve combinatorial optimization by mapping computational problems, such as the factorization of integers in the minimization of many-body Hamiltonians. New photonic technologies involve the use of large-scale optical systems, in which spins are the polarizations of photons in a laser beam. Through the use of new optical devices and nonlinear apparatuses controlled by artificial intelligence, they will prove to be ultra-fast optical computers on the scale of millions of spins.

Neuromorphic computation is an emerging paradigm in the context of new models of neural networks. The problem that today limits the hardware of neural networks is the cost of training these models (used, for example, for language translation or for tracking objects in 3D environments). The required energy and environmental impact of training large neural networks for complex activities is the main difficulty. If we consider the current consumption of tens of billions of kWh worldwide by large data-processing centers, we understand how it is increasingly important to develop new calculation models in which training is not so intense and onerous. New computational models and new, more efficient hardware need to be identified. A paradigm that is emerging is the so-called “neuromorphic” calculation, inspired by the efficient functioning of the human brain. In neuromorphic networks, most of the weights are not optimized, and training takes place only in the input and output if the network is large enough, it can be shown that the computational potentials are comparable to standard models. Furthermore, the neuromorphic schemes are directly implementable with photonic hardware in which light replaces the electronic current for processing. In this optical hardware, power consumption is drastically reduced because only passive components are used, the processing speed is the maximum possible and the scale of problems currently reaches 106 spins and is set to multiply rapidly in the future. Within the Fermi Center, following some proposals of the participants [Marcucci2020], new classical and quantum photonic neuromorphic calculators will be created. It is a new class of experiments and devices that opens up many theoretical and application challenges.

The idea is to create the first link, via fibers or open air, of quantum information that links the original institute in Via Panisperna with the Physics Department of Sapienza. It is a revolutionary experiment potentially with great historical impact, which can also represent the first step towards a quantum network on a larger scale. Furthermore, this link can be integrated with computational machines to test the role of entanglement and quantum non-locality, also in the context of studies on the fundamental principles of quantum mechanics.

Quantum technologies have the potential to profoundly influence various aspects of modern society. Relevant examples are the simulation of quantum systems, materials engineering, nanotechnologies and internet commerce. Machine learning is a vibrant area of ​​research that has progressed very rapidly in recent years: Its applications are ubiquitous and range from e-commerce, healthcare, neuroimaging, and particle physics to fundamental science. The purpose of this line of research is therefore to work experimentally on the connection between quantum information and machine learning. Both an integrated hybrid photonic platform and a “bulk” optical platform will be exploited.

1) We will demonstrate that quantum walk-based photonic platforms can be effectively adopted to implement Quantum Machine Learning protocols. Several quantum machine learning demonstrations will be carried out to demonstrate the ability of quantum agents to learn physical characteristics not accessible through classical techniques. As a paradigm of reference, we will exploit the quantum-computing reservoir based on a photonic platform. 2) We will use machine-learning techniques to certify the correct functioning of quantum devices. The certification of quantum devices is an element of fundamental importance, in particular, in a regime where the system solves a classically intractable problem. The ability of machine-learning techniques to deal with large amounts of data, and to find recurring patterns within them, can therefore represent a powerful tool for validating both communication protocols and quantum algorithms [Gior2018, Agre2019].

Causal inference starting from experimental observations appears to be of primary importance in various scientific fields. The theory of causality has in fact become a fundamental tool for a wide range of applications, such as statistics and machine learning, and through these, in genetics, social studies and economics. It has been found that our basic notions of cause and effect are incompatible with quantum phenomena [Chav2017, Pode2019]. Causation theory provides a powerful new tool for addressing quantum-information problems. Recently it has been shown that quantum causality allows the development of new protocols based on fewer constraints, unlike those implemented previously. Relevant examples are the certified generation of purely random numbers [Agre2020] or the generation of quantum effects within a network between different laboratories [Pode 2020]. This line of research is intended to take this activity to the next level, attaining breakthroughs for both theorists and experimentalists on the experimental basis and implications of quantum causality. In particular, the following objectives will be pursued:

1) To analyze the emergence of new types of non-classical behavior in various types of causal networks never before considered. This includes the use of remote connections via optical fiber and/or free-space between two or more nodes.

2) To develop new quantum-information protocols for quantum networks comprising multiple participants.

3) Take advantage of machine-learning techniques for the analysis of data obtained through complex experimental structures.

  • [Pierangeli2019] D. Pierangeli, G.Marcucci, C. Conti, Large-scale photonic Ising machine by spatial light modulation, Phys. Rev. Lett. 122, 213902 (2019)
  • [Marcucci2020] G. Marcucci, D. Pierangeli, C. Conti, Theory of Neuromorphic Computing by Waves, Phys. Rev. Lett. 2020, to be published, arXiv:1912.07044
  • [Agre2019] I. Agresti, N. Viggianiello, F. Flamini, N. Spagnolo, A. Crespi, R. Osellame, N. Wiebe, F. Sciarrino, Pattern recognition techniques for Boson Sampling validation, Phys. Rev. X 9, 011013 (2019)
  • [Abel2018] C. Abellán,et al., Challenging local realism with human choices – The BIG Bell Test Collaboration, Nature 557, pp. 212–216 (2018)
  • [Giord2018] T. Giordani, F. Flamini, M. Pompili, N. Viggianiello, N. Spagnolo, A. Crespi, R. Osellame, N. Wiebe, M. Walschaers, A. Buchleitner, F. Sciarrino. Experimental statistical signature of many-body quantum interference. Nature Photonics 12, 173–178 (2018)
  • [Wang2019] J. Wang, F. Sciarrino, A. Laing, M. G. Thompson, Integrated photonic quantum technologies, Nature Photonics (2019) doi:10.1038/s41566-019-0532-1.
  • [Chav2017] R. Chaves, G. Carvacho, I. Agresti, V. Di Giulio, L. Aolita, S. Giacomini, F. Sciarrino. Quantum violation of an instrumental test. Nature Physics (2017). doi:10.1038/s41567-017-0008-5
  • [Agre2020] I. Agresti, D. Poderini, L. Guerini, M. Mancusi, G. Carvacho, L. Aolita, D. Cavalcanti, R. Chaves, F. Sciarrino, Experimental device-independent certified randomness generation with an instrumental causal structure, Communications Physics volume 3, article number: 110 (2020).
  • [Pode2019] D. Poderini, R. Chaves, I. Agresti, G. Carvacho, F. Sciarrino, Exclusivity graph approach to Instrumental inequalities, Proceedings of Conference on Uncertainty in Artificial Intelligence (UAI), (2019).
  • [Pode2020] D. Poderini, I. Agresti, G. Marchese, E. Polino, T. Giordani, A. Suprano, M. Valeri, G. Milani, N. Spagnolo, G. Carvacho, R. Chaves and F. Sciarrino, Experimental violation of n-locality in a star quantum network, Nature Communications 11, 2467 (2020).
  • Rat/Mouse PINP EIA
  • IDS-iSYS PTH (1-34)
  • Corticosterone EIA
  • Corticosterone HS
  • Urine CartiLaps EIA
  • Rat/Mouse PINP EIA
  • IDS-iSYS PTH (1-34)
  • Corticosterone EIA
  • Corticosterone HS
  • Urine CartiLaps EIA