## List of seminars (2021-2022)

(the list will be constantly updated as new seminars are scheduled)

**Title**: AI for designing quantum experiments**Speaker**: Dr. Alexey Melnkiov (Terra Quantum AG)**When**: Thursday 21/10/21 at 6.30pm__Register here__**Abstract**: Quantum experiments push the envelope of our understanding of fundamental concepts in quantum physics. The designing of modern quantum experiments is difficult and often clashes with human intuition. In my talk, I will address whether a reinforcement learning agent can propose novel quantum experiments. In our works, we answer this question by considering two examples. In the first example, a reinforcement learning agent learns to create high-dimensional entangled multiphoton states. In the second example, our reinforcement learning agent proposes new unintuitive experiments leading to higher Bell-CHSH inequality violations than the best currently known setups. Our findings highlight the possibility that machine learning could have a significantly more creative role in future quantum experiments.

**Title**: A data-driven perspective on quantum matter**Speaker**: Prof. Eliska Greplova (TU Delft)**When**: Wednesday 17/11/21 at 4pm__Register here__**Abstract**: The fields of condensed matter, artificial intelligence and quantum computing have independently experienced a number of breakthroughs in the last decade. In this talk, I am going provide an illustration of their mutually beneficial overlaps through the lens of machine learning techniques. Specifically, I am going to show how Hamiltonian learning insights can bring condensed matter knowledge into the realm of quantum computing. I will illustrate this approach on quantum error correction problems. A condensed matter physics point of view on quantum error correction codes can be also readily translated into physically informed machine learning ansätze for quantum wave functions, and thus contribute to classical simulation of these models. Finally, I am going to discuss how to use quantum computational complexity insights for more successful variational optimisation.

**Title**: Quantum Reinforcement Learning with Quantum Technologies**Speaker**: Prof. Lucas Lamata (Universidad de Sevilla)**When**: Tuesday 14/12/21 at 4pm**Zoom link****Abstract**: I will describe our theoretical work on the field of quantum reinforcement learning and its implementation in quantum technologies. I will also describe an experiment in collaboration with the Hefei quantum photonics group carrying out our previous proposal in their lab. Quantum reinforcement learning is an exciting novel paradigm which may provide advantages in state tomography and quantum learning with respect to previous algorithms.

**Title**: Similarities and pattern identification in materials-science data**Speaker**: Prof. Claudia Draxl (Humboldt-Universität Berlin)**When**: Thursday 20/01/22 at 4pm**Zoom link****Abstract**: In recent years, data-analytics and machine-learning approaches are being applied to various problems of materials research, and high-throughput screening (HTS) is going hand in hand with the establishment of small- and large-scale data collections. These resources allow us finding trends and patterns that cannot be obtained from individual investigations. Moreover, one can search for materials which exhibit features that are similar to those of other materials but are superior with respect to other criteria. Besides finding materials that resemble each other, e.g. in their electronic properties, we can use the same tools for assessing data quality. As such, one can compare the performance of different methodologies for one and the same material or the impact of approximations and computational parameters on calculated properties. We also make use of un unsupervised learning to find trends in the data and rationalize their physical origin. It will also be discussed what the challenges are for building a FAIR data infrastructure, and how we are currently expanding our efforts toward inclusion of experimental data.

**Title**: Equilibrium and non-equilibrium regimes in Restricted Boltzmann Machines**Speaker**: Dr Beatriz Seoane (Universidad Complutense de Madrid)**When**: Wednesday 16/02/22 at 4pm__Register here__**Abstract**: Training Restricted Boltzmann Machines (RBMs) has been challenging for a long time due to the difficulty of computing precisely the log-likelihood gradient. Over the past decades, many works have proposed more or less successful training recipes but without studying the crucial quantity of the problem: the mixing time, i.e. the number of Monte Carlo iterations needed to sample new configurations from a model. In this work, we show that this mixing time plays a crucial role in the dynamics and stability of the trained model, and that RBMs operate in two well-defined regimes, namely equilibrium and out-of-equilibrium, depending on the interplay between this mixing time of the model and the number of steps, k, used to approximate the gradient. We further show empirically that this mixing time increases with the learning, which often implies a transition from one regime to another as soon as k becomes smaller than this time. In particular, we show that using the popular k (persistent) contrastive divergence approaches, with k small, the dynamics of the learned model are extremely slow and often dominated by strong out-of-equilibrium effects. On the contrary, RBMs trained in equilibrium display faster dynamics, and a smooth convergence to dataset-like configurations during the sampling. Finally we discuss how to exploit in practice both regimes depending on the task one aims to fulfill: (i) short k can be used to generate convincing samples in short learning times, (ii) large k (or increasingly large) is needed to learn the correct equilibrium distribution of the RBM. Finally, the existence of these two operational regimes seems to be a general property of energy based models trained via likelihood maximization.

**Title**: Ultrafast photonic reservoir computing: From fundamental properties to real-world applications**Speaker**: Prof. Miguel Cornelles Soriano (Universitat de les Illes Balears)**When**: Thursday 17/03/22 at 4pm__Register here__**Abstract**: As a result of the increasing computing demands of our globally connected society, novel computing paradigms are on the rise. A promising candidate to tackle this challenge is the so-called neuro-inspired computing, which targets the goal of transferring computational mechanisms used by biological brains to hardware systems beyond the Von Neumann architecture. In this talk, I focus on the neuro-inspired concept of reservoir computing. I present a reservoir computing architecture based on a nonlinear photonic system, implemented by a single semiconductor laser with delayed feedback. Via time-multiplexing, a complex nonlinear network can be emulated using only a single physical node. Neuro-inspired computational concepts like reservoir computing can now be implemented in telecommunication-compatible photonic hardware, with high speed and energy efficiency and exhibiting excellent computing performance. This provides interesting perspectives for applications in optical telecommunication systems.

**Title**: AI usage in structural biology**Speaker**: Prof. Andrea Thorn (Universität Hamburg)**When**: Thursday 05/05/22 at 4pm**Abstract**: Machine learning methods, in particular convolutional neural networks, have been applied to a variety of problems in cryo-EM and macromolecular crystallographic structure solution. However, they still have limited acceptance by the community, mainly in areas where they replace repetitive work and allow for easy visual checking, such as particle picking, crystal centering or crystal recognition. One big exception is machine learning based protein fold prediction which is currently revolutionizing the field. Whether we will be able to exploit this potential fully,will depend on the manner in which we use machine learning: training data must be well-formulated, methods need to utilize appropriate architectures, and outputs must be critically assessed, which may even require explaining artificial intelligence decisions. In this talk, an overview of current applications of machine learning in structural biology will be given, how experimentalist may use fold prediction methods, and it will be discussed what could come next.

**Title**: TBA**Speaker**: Prof. Rubén Pérez (Universidad Autónoma de Madrid)**When**: Wednesday 15/06/22 at 4pm**Abstract**: TBA