Periodically time-varying media, known as photonic time crystals (PTCs), provide a promising platform for observing unconventional wave phenomena. We analyze the scattering of electromagnetic waves from spatially finite PTCs using the multispectral Floquet scattering matrix, which naturally incorporates the frequency-mixing processes intrinsic to such systems. For dispersionless, real, and time-periodic permittivities, this matrix is pseudounitary. Here we demonstrate that this property leads to multiple symmetry-breaking transitions: for increasing driving strength, scattering matrix eigenvalues lying on the unit circle (unbroken symmetry regime) meet at exceptional points (EPs), where they break up into inverse complex conjugate pairs (broken symmetry regime). We identify the symmetry operator associated with these transitions and show that, in time-symmetric systems, it corresponds to the time-reversal operator. Remarkably, at the parametric resonance condition, one eigenvalue vanishes while its partner diverges, signifying simultaneous coherent perfect absorption (CPA) and lasing. Since our approach relies solely on the Floquet scattering matrix, it is not restricted to a specific geometry but instead applies to any periodically time-varying scattering system. To illustrate this universality, we apply our method to a variety of periodically time-modulated structures, including slabs, spheres, and metasurfaces. In particular, we show that using quasi-bound states in the continuum resonances sustained by a metasurface, the CPA and lasing conditions can be attained for a minimal modulation strength of the permittivity. Our results pave the way for engineering time-modulated photonic systems with tailored scattering properties, opening new avenues for dynamic control of light in next-generation optical devices.

We investigate the consequences of information exchange between a system and a measurement-feedback apparatus that cools the system below the environmental temperature. A quantitative relationship between entropy pumping and information acquired about the system is derived, showing that, independent of the concrete realization of the feedback, the latter exceeds the former by a positive amount of excess information flow. This excess information flow satisfies a trade-off relation with the precision of the feedback force, which places strong constraints on both the information-theoretic cost of feedback cooling and the required magnitude of the feedback force. From these constraints, a fundamental lower bound on the energetic cost of optical feedback cooling is derived. Finally, the results are demonstrated for feedback cooling by coherent light scattering. We show that measurement precision is the major factor determining the attainable temperature. Precise measurements can also be leveraged to reduce the required feedback force, leading to significantly more energy-efficient cooling close to the fundamental bound for realistic parameter values.

Visual Question Answering (VQA) is a challenging problem that requires to process multimodal input. Answer-Set Programming (ASP) has shown great potential in this regard to add interpretability and explainability to modular VQA architectures. In this work, we address the problem of how to integrate ASP with modules for vision and natural language processing to solve a new and demanding VQA variant that is concerned with images of graphs (not graphs in symbolic form). Images containing graph-based structures are an ubiquitous and popular form of visualisation. Here, we deal with the particular problem of graphs inspired by transit networks, and we introduce a novel dataset that amends an existing one by adding images of graphs that resemble metro lines. Our modular neuro-symbolic approach combines optical graph recognition for graph parsing, a pretrained optical character recognition neural network for parsing labels, Large Language Models (LLMs) for language processing, and ASP for reasoning. This method serves as a first baseline and achieves an overall average accuracy of 73% on the dataset. Our evaluation provides further evidence of the potential of modular neuro-symbolic systems, in particular with pretrained models that do not involve any further training and logic programming for reasoning, to solve complex VQA tasks.

Quantum squeezed states offer metrological enhancement as compared to their classical counterparts. Here, we devise and numerically explore a novel method for performing SU(1,1) interferometry beyond the standard quantum limit, using quasi-cyclic nonlinear wave mixing dynamics of ultracold atoms in a ring cavity. The method is based on generating quantum correlations between many atoms via photon mediated optomechanical interaction. Timescales of the interferometer operation are here given by the inverse of photonic recoil frequency, and are orders of magnitude shorter than the timescales of collisional spin-mixing based interferometers. Such shorter timescales should enable not only faster measurement cycles, but also lower atomic losses from the trap during measurement, which may lead to significant quantum metrological gain of matter wave interferometry in state of the art cavity setups.

Radiation forces and torques are key to manipulating objects with acoustic or electromagnetic waves. An important concept in this context is the Generalized Wigner-Smith (GWS) matrix, which has previously been primarily studied for optimizing radiation forces and torques on single objects embedded inside complex scattering environments. Here, we develop a unified scattering framework that rigorously establishes this connection for arbitrary inhomogeneous, lossless electromagnetic and acoustic media, as well as for controlling multiple objects individually. Variational identities relate parametric changes of the medium to perturbations of the scattering fields, from which the GWS matrix emerges as a natural generator of radiation forces and torques. For a single object, its extremal eigenstates yield maximal force or torque along a chosen direction; for multiple objects, the same framework defines Pareto-optimal compromises among competing objectives and reveals uncertainty relations for their simultaneous optimization. This establishes a comprehensive foundation towards collective and selective manipulation of objects in complex media.

Adversarial examples have emerged as a significant threat to machine learning algorithms, especially to the convolutional neural networks (CNNs). In this paper, we propose two quantization-based defense mechanisms, Constant Quantization (CQ) and Trainable Quantization (TQ), to increase the robustness of CNNs against adversarial examples. CQ quantizes input pixel intensities based on a "fixed" number of quantization levels, while in TQ, the quantization levels are "iteratively learned during the training phase", thereby providing a stronger defense mechanism. We apply the proposed techniques on undefended CNNs against different state-of-the-art adversarial attacks from the open-source \textit{Cleverhans} library. The experimental results demonstrate 50%-96% and 10%-50% increase in the classification accuracy of the perturbed images generated from the MNIST and the CIFAR-10 datasets, respectively, on commonly used CNN (Conv2D(64, 8x8) - Conv2D(128, 6x6) - Conv2D(128, 5x5) - Dense(10) - Softmax()) available in \textit{Cleverhans} library.

Quantum squeezed states offer metrological enhancement as compared to their classical counterparts. Here, we devise and numerically explore a novel method for performing SU(1,1) interferometry beyond the standard quantum limit, using quasi-cyclic nonlinear wave mixing dynamics of ultracold atoms in a ring cavity. The method is based on generating quantum correlations between many atoms via photon mediated optomechanical interaction. Timescales of the interferometer operation are here given by the inverse of photonic recoil frequency, and are orders of magnitude shorter than the timescales of collisional spin-mixing based interferometers. Such shorter timescales should enable not only faster measurement cycles, but also lower atomic losses from the trap during measurement, which may lead to significant quantum metrological gain of matter wave interferometry in state of the art cavity setups.

For multi-level systems in hot atomic vapors the interplay between the Doppler shift due to atom velocity and the wavenubmer mismatch between driving laser fields strongly influences transmission and absorption properties of the atomic medium. In a three-level atomic ladder-system, Doppler broadening limits the visibility of electromagnetically-induced transparency (EIT) when the probe and control fields are co-propagating, while EIT is recovered under the opposite condition of counter-propagating geometry and k_{p} < k_{c}, with $k_{p}$ and $k_{c}$ being the wavenumbers of the probe and control fields, respectively. This effect has been studied and experimentally demonstrated as an efficient mechanism to realize non-reciprocal probe light transmission, opening promising avenues for example for realization of magnetic-field free optical isolators. In this tutorial we discuss the theoretical derivation of this effect and show the underlying mechanism to be an avoided crossing of the states dressed by the coupling laser as a function of atomic velocities when $k_{p}

The state-of-the-art accelerators for Convolutional Neural Networks (CNNs) typically focus on accelerating only the convolutional layers, but do not prioritize the fully-connected layers much. Hence, they lack a synergistic optimization of the hardware architecture and diverse dataflows for the complete CNN design, which can provide a higher potential for performance/energy efficiency. Towards this, we propose a novel Massively-Parallel Neural Array (MPNA) accelerator that integrates two heterogeneous systolic arrays and respective highly-optimized dataflow patterns to jointly accelerate both the convolutional (CONV) and the fully-connected (FC) layers. Besides fully-exploiting the available off-chip memory bandwidth, these optimized dataflows enable high data-reuse of all the data types (i.e., weights, input and output activations), and thereby enable our MPNA to achieve high energy savings. We synthesized our MPNA architecture using the ASIC design flow for a 28nm technology, and performed functional and timing validation using multiple real-world complex CNNs. MPNA achieves 149.7GOPS/W at 280MHz and consumes 239mW. Experimental results show that our MPNA architecture provides 1.7x overall performance improvement compared to state-of-the-art accelerator, and 51% energy saving compared to the baseline architecture.

We consider the utility maximization problem under convex constraints with regard to theoretical results which allow the formulation of algorithmic solvers which make use of deep learning techniques. In particular for the case of random coefficients, we prove a stochastic maximum principle (SMP), which also holds for utility functions $U$ with $\mathrm{id}_{\mathbb{R}^{+}} \cdot U'$ being not necessarily nonincreasing, like the power utility functions, thereby generalizing the SMP proved by Li and Zheng (2018). We use this SMP together with the strong duality property for defining a new algorithm, which we call deep primal SMP algorithm. Numerical examples illustrate the effectiveness of the proposed algorithm - in particular for higher-dimensional problems and problems with random coefficients, which are either path dependent or satisfy their own SDEs. Moreover, our numerical experiments for constrained problems show that the novel deep primal SMP algorithm overcomes the deep SMP algorithm's (see Davey and Zheng (2021)) weakness of erroneously producing the value of the corresponding unconstrained problem. Furthermore, in contrast to the deep controlled 2BSDE algorithm from Davey and Zheng (2021), this algorithm is also applicable to problems with path dependent coefficients. As the deep primal SMP algorithm even yields the most accurate results in many of our studied problems, we can highly recommend its usage. Moreover, we propose a learning procedure based on epochs which improved the results of our algorithm even further. Implementing a semi-recurrent network architecture for the control process turned out to be also a valuable advancement.

Continuous time crystals, i.e., nonequilibrium phases with a spontaneously broken continuous time-translational symmetry, have been studied and recently observed in the long-time dynamics of open quantum systems. Here, we investigate a lattice of interacting three-level particles and find two distinct time-crystal phases that cannot be described within mean-field theory. Remarkably, one of them emerges only in the presence of correlations, upon accounting for beyond-mean-field effects. Our findings extend explorations of continuous time-translational symmetry breaking in dissipative systems beyond the classical phenomenology of periodic orbits in a low-dimensional nonlinear system. The proposed model applies directly to the laser-driven dynamics of interacting Rydberg states in neutral atom arrays and suggests that the predicted time-crystal phases are observable in such experiments.

Electronic matter waves traveling through the weak and smoothly varying disorder potential of a semi-conductor show branching behavior instead of a smooth spreading of flow. By transferring this phenomenon to optics, we show how the branched flow of light can be controlled to propagate along a single branch rather than through many of them at the same time. Our method is based on shaping the incoming wavefront and only requires partial knowledge of the system's transmission matrix. We show that the light flowing along a single branch has a broadband frequency stability such that we can even steer pulses along selected branches - a prospect with many interesting possibilities for wave control in disordered environments.

Artificial neural networks have become important tools to harness the complexity of disordered or random photonic systems. Recent applications include the recovery of information from light that has been scrambled during propagation through a complex scattering medium, especially in the challenging case where the deterministic input-output transmission matrix cannot be measured. This naturally raises the question of what the limit is that information theory imposes on this recovery process, and whether neural networks can actually reach this limit. To answer these questions, we introduce a model-free approach to calculate the Cram\'er-Rao bound, which sets the ultimate precision limit at which artificial neural networks can operate. As an example, we apply this approach in a proof-of-principle experiment using laser light propagating through a disordered medium, evidencing that a convolutional network approaches the ultimate precision limit in the challenging task of localizing a reflective target hidden behind a dynamically-fluctuating scattering medium. The model-free method introduced here is generally applicable to benchmark the performance of any deep-learning microscope, to drive algorithmic developments and to push the precision of metrology and imaging techniques to their ultimate limit.

We introduce a wavefront shaping protocol for focusing inside disordered media based on a generalization of the established Wigner-Smith time-delay operator. The key ingredient for our approach is the scattering (or transmission) matrix of the medium and its derivative with respect to the position of the target one aims to focus on. A specifc experimental realization in the microwave regime is presented showing that the eigenstates of a corresponding operator are sorted by their focusing strength - ranging from strongly focusing on the designated target to completely bypassing it. Our protocol works without optimization or phase-conjugation and we expect it to be particularly attractive for optical imaging in disordered media.

Speckle patterns are inherent features of coherent light propagation through complex media. As a result of interference, they are sensitive to multiple experimental parameters such as the configuration of disorder or the propagating wavelength. Recent developments in wavefront shaping have made it possible to control speckle pattern statistics and correlations, for example using the concept of the transmission matrix. In this article, we address the problem of correlating scattered fields across multiple degrees of freedom. We highlight the common points between the specific techniques already demonstrated, and we propose a general framework based on the singular value decomposition of a linear combination of multiple transmission matrices. Following analytical predictions, we experimentally illustrate the technique on spectral and temporal correlations, and we show that both the amplitude and the phase of the field correlations can be tuned. Our work opens up new perspectives in speckle correlation manipulation, with potential applications in coherent control.

Using waves to explore our environment is a widely used paradigm, ranging from seismology to radar technology, and from bio-medical imaging to precision measurements. In all of these fields, the central aim is to gather as much information as possible about an object of interest by sending a probing wave at it and processing the information delivered back to the detector. Here, we demonstrate that an electromagnetic wave scattered at an object carries locally defined and conserved information about all of the object's constitutive parameters. Specifically, we introduce here the density and flux of Fisher information for very general types of wave fields and identify corresponding sources and sinks of information through which all these new quantities satisfy a fundamental continuity equation. We experimentally verify our theoretical predictions by studying a movable object embedded inside a disordered environment and by measuring the corresponding Fisher information flux at microwave frequencies. Our results provide a new understanding of the generation and propagation of information and open up new possibilities for tracking and designing the flow of information even in complex environments.

Continual learning is essential for all real-world applications, as frozen pre-trained models cannot effectively deal with non-stationary data distributions. The purpose of this study is to review the state-of-the-art methods that allow continuous learning of computational models over time. We primarily focus on the learning algorithms that perform continuous learning in an online fashion from considerably large (or infinite) sequential data and require substantially low computational and memory resources. We critically analyze the key challenges associated with continual learning for autonomous real-world systems and compare current methods in terms of computations, memory, and network/model complexity. We also briefly describe the implementations of continuous learning algorithms under three main autonomous systems, i.e., self-driving vehicles, unmanned aerial vehicles, and urban robots. The learning methods of these autonomous systems and their strengths and limitations are extensively explored in this article.

Subwavelength atomic arrays feature strong light-induced dipole-dipole interactions, resulting in subradiant collective resonances characterized by narrowed linewidths. In this work, we present a sideband cooling scheme for atoms trapped in subwavelength arrays that utilizes these narrow collective resonances. Working in the Lamb-Dicke regime, we derive an effective master equation for the atomic motion by adiabatically eliminating the internal degrees of freedom of the atoms, and validate its prediction with numerical simulations of the full system. Our results demonstrate that subradiant resonances enable the cooling of ensembles of atoms to temperatures lower than those achievable without dipole interactions, provided the atoms have different trap frequencies. Remarkably, narrow collective resonances can be sideband-resolved even when the individual atomic transition is not. In such scenarios, ground-state cooling becomes feasible solely due to light-induced dipole-dipole interactions. This approach could be utilized for future quantum technologies based on dense ensembles of emitters, and paves the way towards harnessing many-body cooperative decay for enhanced motional control.

Structured waves are ubiquitous for all areas of wave physics, both classical and quantum, where the wavefields are inhomogeneous and cannot be approximated by a single plane wave. Even the interference of two plane waves, or a single inhomogeneous (evanescent) wave, provides a number of nontrivial phenomena and additional functionalities as compared to a single plane wave. Complex wavefields with inhomogeneities in the amplitude, phase, and polarization, including topological structures and singularities, underpin modern nanooptics and photonics, yet they are equally important, e.g., for quantum matter waves, acoustics, water waves, etc. Structured waves are crucial in optical and electron microscopy, wave propagation and scattering, imaging, communications, quantum optics, topological and non-Hermitian wave systems, quantum condensed-matter systems, optomechanics, plasmonics and metamaterials, optical and acoustic manipulation, and so forth. This Roadmap is written collectively by prominent researchers and aims to survey the role of structured waves in various areas of wave physics. Providing background, current research, and anticipating future developments, it will be of interest to a wide cross-disciplinary audience.