FAU Profile Center Light.Matter.QuantumTechnologies

Right now, we are at the Annual Meeting of the #MaxPlanckCentre for Extreme and Quantum Photonics in Ottawa. Great to hear about so many exciting developments in nonlinear optics, quantum optics, attosecond science and more. http://www.mpc-eqp.ca/en/ Max Planck Institute for the Science of Light University of Ottawa Max Planck Society Robert Boyd Gerd Leuchs Vahid Sandoghdar Paul Corkum Ebrahim Karimi Hanieh Fattahi Birgit Stiller Nicolas Joly Jeff Lundeen (and many more)

Making machine learning more sustainable: Neural networks made of light   Dr Clara Wanjura and Prof Florian Marquardt propose a new way of implementing a neural network with an optical system which could make machine learning more sustainable in the future. The researchers at the Max Planck Institute for the Science of Light have published their new method in Nature Physics, demonstrating a method much simpler than previous approaches. Learn more about their inspiring approach 👉 https://lnkd.in/dzxs8JhE #neuromorphic #computing #optics #photonics #machinelearning 📸 © CC Wanjura

When at a Bavarian "Wirtshaus", have you ever heard people at a table call out “Wenz”, “Solo” or “Sie”? If so, you were witnessing a game of “#Schafkopf”, a traditional card game which has now been spiced up with elements out of #QuantumPhysics by #FAUresearcher Ludwig Nützel from the Chair of Theoretical Physics II. Now, how could that look like? Find out in this #interview with Nützel. ⬇ https://lnkd.in/euRFcrMC

It was a great pleasure to contribute a talk today at the innovation day of the SAOT - Erlangen Graduate School in Advanced Optical Technologies at the FAU Erlangen-Nürnberg. My talk was entitled “Quantum Science of Correlated Light and Matter” and covered some of our activities within the collaborative research center QuCoLiMa and the FAU Profile Center Light.Matter.QuantumTechnologies

Deep neural networks can learning the dynamics of physical observables from measurements and even extrapolate their predictions to larger system sizes and longer evolution times. These capacities however strongly depend on whether the dynamical scenario features many body localization or is a scrambling regime. Happy to see this out in Quantum - the open journal for quantum science: https://lnkd.in/dcMpjjZE Thanks to Naeimeh Mohseni, and Tim Byrnes FAU Profile Center Light.Matter.QuantumTechnologies, FAU Erlangen-Nürnberg, Munich Quantum Valley

Our new preprint is out: https://lnkd.in/eSNVVZZy We designed the first Quantum Convolutional Neural Network for a 2-dimensional system. In this work, we show that our network is capable of recognizing intrinsic topological order on the Toric Code up to a specific noise threshold. Big thanks to Nathan McMahon, Petr Zapletal, Hartmann Michael and everyone involved.

For all friends of #neuromorphic_computing: we are happy that our article with Clara Wanjura has just been published in #NaturePhysics ("Fully nonlinear neuromorphic computing with linear wave scattering"). We show how to produce nonlinear information processing in any tuneable linear optical device and how to train such devices very efficiently in a physics-based way. Since tuneable linear optical devices are extremely widespread, these advances should enable neuromorphic computing in many new platforms, promising an energy-efficient alternative to digital neural networks. Usually, linear optical devices, even when highly tuneable, are not very powerful neuromorphic platforms, since their expressivity is limited. After all, they can only represent linear functions of the input, possibly with a single nonlinearity at the end arising from the detection process. However, injecting the input via tuneable parameters gets rid of this restriction and we observe that this approach can give results at the level of digital artificial neural networks. Importantly, we show that obtaining the gradients needed for training is easy: it can be achieved using a small number of scattering experiments, independent of the number of trainable parameters. This is as efficient as the backpropagation algorithm, but now implemented in a physics-based way. Such physics-based training approaches are highly sought-after but still rare. If you have linear optical devices that you think might benefit from this approach, we would be happy to hear from you! Max Planck Institute for the Science of Light Read the paper: https://lnkd.in/eNppVgY7 Read the News&Views by Peter McMahon: https://lnkd.in/eGb4SNUP Press release for the general public: https://lnkd.in/eDfjYbUN #MachineLearning #ArtificialIntelligence #NeuromorphicComputing Picture credit: Clara Wanjura