Automated machine learning (AutoML) supports the algorithmic construction and data-specific customization of machine learning pipelines, including the selection, combination, and parametrization of machine learning algorithms as main constituents. Generally speaking, AutoML approaches comprise two major components: a search space model and an optimizer for traversing the space. Recent approaches have shown impressive results in the realm of supervised learning, most notably (single-label) classification (SLC). Moreover, first attempts at extending these approaches towards multi-label classification (MLC) have been made. While the space of candidate pipelines is already huge in SLC, the complexity of the search space is raised to an even higher power in MLC. One may wonder, therefore, whether and to what extent optimizers established for SLC can scale to this increased complexity, and how they compare to each other. This paper makes the following contributions: First, we survey existing approaches to AutoML for MLC. Second, we augment these approaches with optimizers not previously tried for MLC. Third, we propose a benchmarking framework that supports a fair and systematic comparison. Fourth, we conduct an extensive experimental study, evaluating the methods on a suite of MLC problems. We find a grammar-based best-first search to compare favorably to other optimizers.

Automated Machine Learning (AutoML) seeks to automatically find so-called machine learning pipelines that maximize the prediction performance when being used to train a model on a given dataset. One of the main and yet open challenges in AutoML is an effective use of computational resources: An AutoML process involves the evaluation of many candidate pipelines, which are costly but often ineffective because they are canceled due to a timeout. In this paper, we present an approach to predict the runtime of two-step machine learning pipelines with up to one pre-processor, which can be used to anticipate whether or not a pipeline will time out. Separate runtime models are trained offline for each algorithm that may be used in a pipeline, and an overall prediction is derived from these models. We empirically show that the approach increases successful evaluations made by an AutoML tool while preserving or even improving on the previously best solutions.

Simple thermal treatment of guanine at temperatures ranging from 600 to 700 °C leads to C1N1 condensates with unprecedented CO2/N2 selectivity when compared to other carbonaceous solid sorbents. Increasing the surface area of the CN condensates in the presence of ZnCl2 salt melts enhances the amount of CO2 adsorbed while preserving the high selectivity values and C1N1 structure. Results indicate that these new materials show a sorption mechanism a step closer to that of natural CO2 caption proteins and based on metal free structural cryptopores.

Model predictive control is a prominent approach to construct a feedback control loop for dynamical systems. Due to real-time constraints, the major challenge in MPC is to solve model-based optimal control problems in a very short amount of time. For linear-quadratic problems, Bemporad et al. have proposed an explicit formulation where the underlying optimization problems are solved a priori in an offline phase. In this article, we present an extension of this concept in two significant ways. We consider nonlinear problems and - more importantly - problems with multiple conflicting objective functions. In the offline phase, we build a library of Pareto optimal solutions from which we then obtain a valid compromise solution in the online phase according to a decision maker's preference. Since the standard multi-parametric programming approach is no longer valid in this situation, we instead use interpolation between different entries of the library. To reduce the number of problems that have to be solved in the offline phase, we exploit symmetries in the dynamical system and the corresponding multiobjective optimal control problem. The results are verified using two different examples from autonomous driving.

The machine recognition of speech spoken at a distance from the microphones, known as far-field automatic speech recognition (ASR), has received a significant increase of attention in science and industry, which caused or was caused by an equally significant improvement in recognition accuracy. Meanwhile it has entered the consumer market with digital home assistants with a spoken language interface being its most prominent application. Speech recorded at a distance is affected by various acoustic distortions and, consequently, quite different processing pipelines have emerged compared to ASR for close-talk speech. A signal enhancement front-end for dereverberation, source separation and acoustic beamforming is employed to clean up the speech, and the back-end ASR engine is robustified by multi-condition training and adaptation. We will also describe the so-called end-to-end approach to ASR, which is a new promising architecture that has recently been extended to the far-field scenario. This tutorial article gives an account of the algorithms used to enable accurate speech recognition from a distance, and it will be seen that, although deep learning has a significant share in the technological breakthroughs, a clever combination with traditional signal processing can lead to surprisingly effective solutions.

The control of complex systems is of critical importance in many branches of science, engineering, and industry, many of which are governed by nonlinear partial differential equations. Controlling an unsteady fluid flow is particularly important, as flow control is a key enabler for technologies in energy (e.g., wind, tidal, and combustion), transportation (e.g., planes, trains, and automobiles), security (e.g., tracking airborne contamination), and health (e.g., artificial hearts and artificial respiration). However, the high-dimensional, nonlinear, and multi-scale dynamics make real-time feedback control infeasible. Fortunately, these high- dimensional systems exhibit dominant, low-dimensional patterns of activity that can be exploited for effective control in the sense that knowledge of the entire state of a system is not required. Advances in machine learning have the potential to revolutionize flow control given its ability to extract principled, low-rank feature spaces characterizing such complex systems.We present a novel deep learning modelpredictive control framework that exploits low-rank features of the flow in order to achieve considerable improvements to control performance. Instead of predicting the entire fluid state, we use a recurrent neural network (RNN) to accurately predict the control relevant quantities of the system, which are then embedded into an MPC framework to construct a feedback loop. In order to lower the data requirements and to improve the prediction accuracy and thus the control performance, incoming sensor data are used to update the RNN online. The results are validated using varying fluid flow examples of increasing complexity.

<jats:p>In the present work, we provide an electronic structure based method for the “on-the-fly” determination of vibrational sum frequency generation (v-SFG) spectra. The predictive power of this scheme is demonstrated at the air-water interface. While the instantaneous fluctuations in dipole moment are obtained using the maximally localized Wannier functions, the fluctuations in polarizability are approximated to be proportional to the second moment of Wannier functions. The spectrum henceforth obtained captures the signatures of hydrogen bond stretching, bending, as well as low-frequency librational modes.</jats:p>

Photovoltaics is one of the most promising and fastest-growing renewable energy technologies. Although the price-performance ratio of solar cells has improved significantly over recent years{,} further systematic investigations are needed to achieve higher performance and lower cost for future solar cells. In conjunction with experiments{,} computer simulations are powerful tools to investigate the thermodynamics and kinetics of solar cells. Over the last few years{,} we have developed and employed advanced computational techniques to gain a better understanding of solar cells based on copper indium gallium selenide (Cu(In{,}Ga)Se2). Furthermore{,} we have utilized state-of-the-art data-driven science and machine learning for the development of photovoltaic materials. In this Perspective{,} we review our results along with a survey of the field.

<jats:p>The three-dimensional (3D) crystal structures of the GAF3 domain of cyanobacteriochrome Slr1393 (<jats:italic>Synechocystis</jats:italic> PCC6803) carrying a phycocyanobilin chromophore could be solved in both 15-<jats:italic>Z</jats:italic> dark-adapted state, Pr, λ<jats:sub>max</jats:sub> = 649 nm, and 15-<jats:italic>E</jats:italic> photoproduct, Pg, λ<jats:sub>max</jats:sub> = 536 nm (resolution, 1.6 and 1.86 Å, respectively). The structural data allowed identifying the large spectral shift of the Pr-to-Pg conversion as resulting from an out-of-plane rotation of the chromophore’s peripheral rings and an outward movement of a short helix formed from a formerly unstructured loop. In addition, a third structure (2.1-Å resolution) starting from the photoproduct crystals allowed identification of elements that regulate the absorption maxima. In this peculiar form, generated during X-ray exposition, protein and chromophore conformation still resemble the photoproduct state, except for the D-ring already in 15-<jats:italic>Z</jats:italic> configuration and tilted out of plane akin the dark state. Due to its formation from the photoproduct, it might be considered an early conformational change initiating the parental state-recovering photocycle. The high quality and the distinct features of the three forms allowed for applying quantum-chemical calculations in the framework of multiscale modeling to rationalize the absorption maxima changes. A systematic analysis of the PCB chromophore in the presence and absence of the protein environment showed that the direct electrostatic effect is negligible on the spectral tuning. However, the protein forces the outer pyrrole rings of the chromophore to deviate from coplanarity, which is identified as the dominating factor for the color regulation.</jats:p>

CP2K is an open source electronic structure and molecular dynamics software package to perform atomistic simulations of solid-state, liquid, molecular, and biological systems. It is especially aimed at massively parallel and linear-scaling electronic structure methods and state-of-theart ab initio molecular dynamics simulations. Excellent performance for electronic structure calculations is achieved using novel algorithms implemented for modern high-performance computing systems. This review revisits the main capabilities of CP2K to perform efficient and accurate electronic structure simulations. The emphasis is put on density functional theory and multiple post–Hartree–Fock methods using the Gaussian and plane wave approach and its augmented all-electron extension.

Polarons in dielectric crystals play a crucial role for applications in integrated electronics and optoelectronics. In this work, we use density-functional theory and Green's function methods to explore the microscopic structure and spectroscopic signatures of electron polarons in lithium niobate (LiNbO3). Total-energy calculations and the comparison of calculated electron paramagnetic resonance data with available measurements reveal the formation of bound polarons at Nb_Li antisite defects with a quasi-Jahn-Teller distorted, tilted configuration. The defect-formation energies further indicate that (bi)polarons may form not only at Nb_Li antisites but also at structures where the antisite Nb atom moves into a neighboring empty oxygen octahedron. Based on these structure models, and on the calculated charge-transition levels and potential-energy barriers, we propose two mechanisms for the optical and thermal splitting of bipolarons, which provide a natural explanation for the reported two-path recombination of bipolarons. Optical-response calculations based on the Bethe-Salpeter equation, in combination with available experimental data and new measurements of the optical absorption spectrum, further corroborate the geometries proposed here for free and defect-bound (bi)polarons.

We numerically simulate multiple light scattering in discrete disordered media represented by large clusters of irregular non-absorbing particles. The packing density of clusters is 0.5. With such conditions diffuse scattering is significantly reduced and light transport follows propagation channels that are determined by the particle size and topology of the medium. This kind of localization produces coherent backscattering intensity surge and enhanced negative polarization branch if compared to lower density samples.

Individual grains of chalcopyrite solar cell absorbers can facet in different crystallographic directions at their surfaces. To gain a deeper understanding of the junction formation in these devices, we correlate variations in the surface facet orientation with the defect electronic properties. We use a combined analytical approach based on scanning tunneling spectroscopy (STS), scanning electron microscopy, and electron back scatter diffraction (EBSD), where we perform these experiments on identical surface areas as small as 2 × 2 µm2 with a lateral resolution well below 50 nm. The topography of the absorber surfaces indicates two main morphological features: micro-faceted, long basalt-like columns and their short nano-faceted terminations. Our STS results reveal that the long columns exhibit spectral signatures typical for the presence of pronounced oxidation-induced surface dipoles in conjunction with an increased density of electronic defect levels. In contrast, the nano-faceted terminations of the basalt-like columns are largely passivated in terms of electronic defect levels within the band gap region. Corresponding crystallographic data based on EBSD experiments show that the surface of the basalt-like columns can be assigned to intrinsically polar facet orientations, while the passivated terminations are assigned to non-polar planes. Ab-initio calculations suggest that the polar surfaces are more prone to oxidation and resulting O-induced defects, in comparison to non-polar planes. Our results emphasize the correlation between morphology, surface facet orientations and surface electronic properties. Furthermore, this work aids in gaining a fundamental understanding of oxidation induced lateral inhomogeneities in view of the p-n junction formation in chalcopyrite thin-film solar cells.

The electrochemical nitrogen reduction reaction (NRR) to ammonia (NH3) is a promising alternative route for an NH3 synthesis at ambient conditions to the conventional high temperature and pressure Haber--Bosch process without the need for hydrogen gas. Single metal ions or atoms are attractive candidates for the catalytic activation of non-reactive nitrogen (N2), and for future targeted improvement of NRR catalysts, it is of utmost importance to get detailed insights into structure-performance relationships and mechanisms of N2 activation in such structures. Here, we report density functional theory studies on the NRR catalyzed by single Au and Fe atoms supported in graphitic C2N materials. Our results show that the metal atoms present in the structure of C2N are the reactive sites, which catalyze the aforesaid reaction by strong adsorption and activation of N2. We further demonstrate that a lower onset electrode potential is required for Fe--C2N than for Au--C2N. Thus, Fe--C2N is theoretically predicted to be a potentially better NRR catalyst at ambient conditions than Au--C2N owing to the larger adsorption energy of N2 molecules. Furthermore, we have experimentally shown that single sites of Au and Fe supported on nitrogen-doped porous carbon are indeed active NRR catalysts. However, in contrast to our theoretical results, the Au-based catalyst performed slightly better with a Faradaic efficiency (FE) of 10.1{\%} than the Fe-based catalyst with an FE of 8.4{\%} at −0.2 V vs. RHE. The DFT calculations suggest that this difference is due to the competitive hydrogen evolution reaction and higher desorption energy of ammonia.

In scientific computing, the acceleration of atomistic computer simulations by means of custom hardware is finding ever-growing application. A major limitation, however, is that the high efficiency in terms of performance and low power consumption entails the massive usage of low precision computing units. Here, based on the approximate computing paradigm, we present an algorithmic method to compensate for numerical inaccuracies due to low accuracy arithmetic operations rigorously, yet still obtaining exact expectation values using a properly modified Langevin-type equation.

<p>EPR spectroscopy reveals the universality class and dynamic effects of the [NH₄][Zn(HCOO)₃] hybrid formate framework.</p>

The challenge of designing new tunable nonlinear dielectric materials with tailored properties has attracted an increasing amount of interest recently. Herein, we study the effective nonlinear dielectric response of a stochastic paraelectric-dielectric composite consisting of equilibrium distributions of circular and partially penetrable disks (or parallel, infinitely long, identical, partially penetrable, circular cylinders) of a dielectric phase randomly dispersed in a continuous matrix of a paraelectric phase. The random microstructures were generated using the Metropolis Monte Carlo algorithm. The evaluation of the effective permittivity and tunability were carried out by employing either a Landau thermodynamic model or its Johnson’s approximation to describe the field-dependent permittivity of the paraelectric phase and solving continuum-electrostatics equations using finite element calculations. We reveal that the percolation threshold in this composite governs the critical behavior of the effective permittivity and tunability. For microstructures below the percolation threshold, our simulations demonstrate a strong nonlinear behaviour of the field-dependent effective permittivity and very high tunability that increases as a function of dielectric phase concentration. Above the percolation threshold, the effective permittivity shows the tendency to linearization and the tunability dramatically drops down. The highly reduced permittivity and extraordinarily high tunability are obtained for the composites with dielectric impenetrable disks at high concentrations, in which the triggering of the percolation transition is avoided. The reported results cast light on distinct nonlinear behaviour of 2D and 3D stochastic composites and can guide the design of novel composites with the controlled morphology and tailored permittivity and tunability.

A stepwise angular spectrum method (SASM) for curved interfaces is presented to calculate the wave propagation in planar lens-like integrated optical structures based on photonic slab waveguides. The method is derived and illustrated for an effective 2D setup first and then for 3D slab waveguide lenses. We employ slab waveguides of different thicknesses connected by curved surfaces to realize a lens-like structure. To simulate the wave propagation in 3D including reflection and scattering losses, the stepwise angular spectrum method is combined with full vectorial finite element computations for subproblems with lower complexity. Our SASM results show excellent agreement with rigorous numerical simulations of the full structures with a substantially lower computational effort and can be utilized for the simulation-based design and optimization of complex and large scale setups.

This is the second part of a project on the foundations of first-principle calculations of the electron transport in crystals at finite temperatures, aiming at a predictive first-principles platform that combines ab-initio molecular dynamics (AIMD) and a finite-temperature Kubo-formula with dissipation for thermally disordered crystalline phases. The latter are encoded in an ergodic dynamical system (Ω,G,dP), where Ω is the configuration space of the atomic degrees of freedom, G is the space group acting on Ω and dP is the ergodic Gibbs measure relative to the G-action. We first demonstrate how to pass from the continuum Kohn–Sham theory to a discrete atomic-orbitals based formalism without breaking the covariance of the physical observables w.r.t. (Ω,G,dP). Then we show how to implement the Kubo-formula, investigate its self-averaging property and derive an optimal finite-volume approximation for it. We also describe a numerical innovation that made possible AIMD simulations with longer orbits and elaborate on the details of our simulations. Lastly, we present numerical results on the transport coefficients of crystal silicon at different temperatures.

<p>The hard X-ray spectroscopy methods XAS, valence-to-core XES and higher solution XANES offer unique insights into organometallic reaction mechanisms.</p>

Existing approaches and tools for the generation of approximate circuits often lack generality and are restricted to certain circuit types, approximation techniques, and quality assurance methods. Moreover, only few tools are publicly available. This hinders the development and evaluation of new techniques for approximating circuits and their comparison to previous approaches. In this paper, we first analyze and classify related approaches and then present CIRCA, our flexible framework for search-based approximate circuit generation. CIRCA is developed with a focus on modularity and extensibility. We present the architecture of CIRCA with its clear separation into stages and functional blocks, report on the current prototype, and show initial experiments.

<jats:p>Cyanobacteriochromes are compact and spectrally diverse photoreceptor proteins that are promising candidates for biotechnological applications. Computational studies can contribute to an understanding at a molecular level of their wide spectral tuning and diversity. In this contribution, we benchmark methods to model a 110 nm shift in the UV/Vis absorption spectrum from a red- to a green-absorbing form of the cyanobacteriochrome Slr1393g3. Based on an assessment of semiempirical methods to describe the chromophore geometries of both forms in vacuo, we find that DFTB2+D leads to structures that are the closest to the reference method. The benchmark of the excited state calculations is based on snapshots from quantum mechanics/molecular mechanics molecular dynamics simulations. In our case, the methods RI-ADC(2) and sTD-DFT based on CAM-B3LYP ground state calculations perform the best, whereas no functional can be recommended to simulate the absorption spectra of both forms with time-dependent density functional theory. Furthermore, the difference in absorption for the lowest energy absorption maxima of both forms can already be modelled with optimized structures, but sampling is required to improve the shape of the absorption bands of both forms, in particular for the second band. This benchmark study can guide further computational studies, as it assesses essential components of a protocol to model the spectral tuning of both cyanobacteriochromes and the related phytochromes.</jats:p>