Quantum machine learning in simulated linear-optics scenarios
Outline, including interdisciplinary dimension
Learning is the process of acquiring, modifying, and recognising knowledge and involves the synthesis of different forms of information. Learning is not just a human prerogative but extends to some types of machines, which are able to acquire stimuli and process this information to evolve accordingly. Besides being relevant in designing optimal control techniques in classical artificial intelligence, machine learning (ML) turns out to be adaptable to the quantum domain. In this context, a physical system performs a given task and receives feedback, which is in turn used by the machine itself to change its behaviour, and thus learn.
A key problem in quantum computer science is to design the transduction function that processes the input data received by the (quantum) machine after each feedback step and changes the rules of the machines functioning so to optimise the learning process. It is alleged that quantum ML (qML) can provide an exponential speed-up over classical computers for a variety of learning tasks, including supervised and un-supervised algorithms. However, what is the ultimate reason behind the advantages provided by qML over its classical counterpart? It is believed that quantum coherences are key to the enhanced performance of qML over classical ML. Can we thus design experimental architectures for the former that exploit quantum coherence to implement the algorithms devised for a machine to efficiently learn a task?
The overarching goal of this project is to design experimentally testable architectures for the implementation of test-bed qML algorithms in simulated linear optics scenarios. We will explore theoretically, and simulate experimentally, quantum models able to naturally embody the correct computational resources for the performance of qML algorithms in the adiabatic, supervised and un-supervised scenario. The proven capacity of multi-mode, multi-degree-of-freedom linear optics settings to embody flexible platforms for medium-scale quantum simulators will pave the way to the implementation of test-beds able to prove the significance of the qML paradigm.
The project is inherent interdisciplinary, requiring elements of quantum information, computer science, and experimental/theoretical quantum optics. It is based on the collaboration between the theoretical team at Queen's led by Paternostro, and the experimental one working in linear optics at the Rome labs.
Key words/descriptors
quantum technologies, quantum computer science, experimental linear optics, quantum machine learning
First supervisor
Professor Mauro Paternostro - School of Mathematics and Physics
Second and third supervisors from a complementary discipline
Professor Fabio Sciarrino - Universita' di Roma (Italy)
Professor Paolo Mataloni - Universita' di Roma (Italy)
Supervisors’ track record of PhD completions, plus excellence and international standing in the project area
This is a solid existing collaboration. Paternostro is a frequent collaborator of both Sciarrino and Mataloni, with who has published 6 research papers in the last 4 years (including papers that have appeared in Phys. Rev. Lett. and Nature Scientific Reports, which are among the premier journals in the quantum information processing area). Moreover, they are partners in current bids to the EU Horizon2020 FET call within the framework of a project that address questions closely related to the topic of this project.
Both Paternostro and Sciarrino have a strong track record of supervision of postgraduate students. Since appointment, Paternostro has supervised to completion (as first or second supervisor) 6 PhD students at QUB. He was co-supervisor of a PhD student at University College Cork and is co-supervising a postgraduate student at Nanyang Technological University (Singapore). He is current first supervisor of 4 more students. Sciarrino has supervised or co-supervised 6 PhD students at Sapienza Universita' di Roma. All students supervised by the proposed team have successfully found employment either in academia (5 out of the 6 supervised by Paternostro), finance, or industry.
Both Paternostro and Sciarrino are internationally recognised as leaders in their respective area. The world-class standing of the proposed supervisory team is demonstrated and witnessed by the prestigious grants awarded to both (Paternostro: EPSRC Career acceleration Fellowship, Alexander von Humboldt Fellowship, Marie Curie IEF leadership, Coordination of a EU FP7 grant, PI in a other national and international grants; Sciarrino: ERC Consolidator Grant 2012, Coordination of 1 FP7 grant, 1 Marie Curie ITN, 1 Horizon2020 grant, besides numerous national awards) and 100+ iinvitations to talk at international events that they have received to date. The excellence of their standing is further remarked by their participation to review and evaluation panels at the national and European level (Paternostro: EPSRC College member, reviewer for the Royal Society University Research Fellowship and Newton Fellowship, reviewer for the EU Horizon2020 programme and the Marie Curie ITN; Sciarrino: reviewer for the EU FP7 and the ERC).
Intersectoral exposure and/or international mobility
(e.g. secondments to/collaboration with partner organizations)
The student will spend 3 months per year within the Rome group, working with the Sciarrino and Mataloni's team to demonstrate experimentally the theoretical framework that will be developed in Belfast. Therefore, the international mobility aspects of the project are clearly demonstrated, and the COFUND requests fully met.
Secondments to chosen industrial partners will take place on a yearly basis (for an estimated length of 2 weeks each) and will be aimed at providing the student with important tranferrable skills of a translational nature in explicitly non-academic aspects (project management, risk assessment, marketing and alike). The training programme is thus explicitly intersectoral and interdisciplinary in nature, fostering the collaboration between scientists with complementary skills and expertise, and integrating an important contribution from relevant industrial actors.
Describe briefly the international profile of the partner
The group led by Profs Mataloni and Sciarrino is world leading in the field of experimental linear optics. The Rome labs have performed pioneering work in the quantum-limited management of integrated optics, the harnessing of orbital angular momentum of light for quantum communication, the demonstration of quantum information primitives and the manipulation of multi-qubit quantum states for quantum computation purposes. The international standing of the group is of the highest level, as witnessed by the numerous successes achieved in securing funding from competitive international calls (including CHIST-ERA, the EU FP7 and Horizon2020, ERC). The group offers diversity of interests, deep expertise and a network of international collaborations with some of the premier theoretical and experimental groups worldwide. Both Mataloni and Sciarrino have a strong track record of supervision of postgraduate students and postdoctoral fellows. Some of the current young leaders in the field of experimental quantum optics have studied or worked in the Rome group.
Training that will be provided through the research project itself
The student will acquire an ample set of skills both at the theoretical and experimental level. He will develop expertise in theoretical quantum information processing, quantum computer science and the theory of quantum control. By interacting with the Rome group, he will be exposed to world-class experimental work, thus acquiring knowledge in the difficulties linked to the management of quantum experiments. The student will develop skills in the establishment of a common collaborative language between theorists and experimentalists. Moreover, through secondments at industrial collaborators such as ID-Quantique, QUTOOLS, Adapticam and Single Quantum, which are interested in the development and commercialization of quantum technologies for information processing and in the topics of this project, the student will be trained in fully transferable, inter-disciplinary skills explicitly oriented towards industrial needs in the burgeoning field of quantum technologies.
Examples of additional training in non-research transferable skills
The industrial partner will contribute with training in complementary skills like project management, organizational skills, leadership, entrepreneurship, patent issues, economics and understanding of commerce, risk management, communication skills, team-building etc. in order to ensure excellent training on abilities required in the (quantum technology) market.
Expected dissemination of results: peer-reviewed journals, seminars, workshop and conferences at European/international level
(e.g. public talks, visits to schools, open days, QUB impact showcase)
Results achieved through the development of the proposed project will be disseminated through high-impact peer-reviewed journals such as Phys. Rev. Lett. and Phys. Rev. X, which are leading journals for the theoretical physics area, and journals of the Nature Publishing Group, where the Rome team frequently reports their results.
Moreover, major events explicitly addressing quantum information and quantum computer science such as the Quantum information Processing conference series and the International Conference on Quantum Communication, Measurements, and Computing series will be targeted to present our results. At the European level, we will target the Central European Workshops on Quantum Optics, which is a major event where the main groups working on quantum optics in Europe present their most exciting results.
Expected impact activities
(e.g. public talks, visits to schools, open days, QUB impact showcase)
Both Sciarrino and Paternostro are popular speakers and have a track record of presentations aimed at the general public and more subject-specific audiences. They will use any opportunity in this sense to present the results of the proposed project, both nationally and internationally. Paternostro frequently provides A-level talks on quantum physics to schools in the Belfast area, while Sciarrino has been involved in the organisation and presentation of scientific lectures to the general public in the Rome area. They will continue such activities focusing in particular on the topics of the proposed project, which is well suited to the understanding at the laymen level.
Paternostro will also make use of his role as a career speaker at open days in Applied Mathematics at Queen's to advetise the activities of the University in active research oriented towards the quantum world.