Search

This paper introduces a new biologically-inspired training method named Continual Learning through Adjustment Suppression and Sparsity Promotion (CLASSP). CLASSP is based on two main principles observed in neuroscience, particularly in the context of synaptic transmission and Long-Term Potentiation (LTP). The first principle is a decay rate over the weight adjustment, which is implemented as a generalization of the AdaGrad optimization algorithm. This means that weights that have received many updates should have lower learning rates as they likely encode important information about previously seen data. However, this principle results in a diffuse distribution of updates throughout the model, as it promotes updates for weights that haven't been previously updated, while a sparse update distribution is preferred to leave weights unassigned for future tasks. Therefore, the second principle introduces a threshold on the loss gradient. This promotes sparse learning by updating a weight only if the loss gradient with respect to that weight is above a certain threshold, i.e. only updating weights with a significant impact on the current loss. Both principles reflect phenomena observed in LTP, where a threshold effect and a gradual saturation of potentiation have been observed. CLASSP is implemented in a Python/PyTorch class, making it applicable to any model. When compared with Elastic Weight Consolidation (EWC) using Computer Vision and sentiment analysis datasets, CLASSP demonstrates superior performance in terms of accuracy and memory footprint., Comment: In this version I included a new experiment in text classification using transformer architecture

This work introduces a novel R package for concise, informative summaries of machine learning models. We take inspiration from the summary function for (generalized) linear models in R, but extend it in several directions: First, our summary function is model-agnostic and provides a unified summary output also for non-parametric machine learning models; Second, the summary output is more extensive and customizable -- it comprises information on the dataset, model performance, model complexity, model's estimated feature importances, feature effects, and fairness metrics; Third, models are evaluated based on resampling strategies for unbiased estimates of model performances, feature importances, etc. Overall, the clear, structured output should help to enhance and expedite the model selection process, making it a helpful tool for practitioners and researchers alike., Comment: 9 pages

Supercontinua are broadband spectra that are essential to optical spectroscopy, sensing, imaging, and metrology. They are generated from ultrashort laser pulses through nonlinear frequency conversion in fibers, bulk media, and chip-integrated waveguides. For any generating platform, balancing the competing criteria of strong nonlinearity, transparency, and absence of multiphoton absorption is a key challenge. Here, we explore supercontinuum generation in integrated gallium nitride (GaN) waveguides, which combine a high Kerr-nonlinearity, mid-infrared transparency, and a large bandgap that prevents two- and three-photon absorption in the technologically important telecom C-band, where compact erbium-based pump lasers exist. Using this type of laser, we demonstrate tunable dispersive waves and gap-free spectra extending to almost 4 micron in wavelength, relevant to functional group chemical sensing. Additionally, leveraging the material's second-order nonlinearity, we implement on-chip f-2f interferometry to detect the pump laser's carrier-envelope offset frequency, which enables precision metrology. These results demonstrate the versatility of GaN-on-sapphire as a new platform for broadband nonlinear photonics.

A common goal in clinical trials is to conduct tests on estimated treatment effects adjusted for covariates such as age or sex. Analysis of Covariance (ANCOVA) is often used in these scenarios to test the global null hypothesis of no treatment effect using an $F$-test. However, in several samples, the $F$-test does not provide any information about individual null hypotheses and has strict assumptions such as variance homoscedasticity. We extend the method proposed by Konietschke et al. (2021) to a multiple contrast test procedure (MCTP), which allows us to test arbitrary linear hypotheses and provides information about the global as well as the individual null hypotheses. Further, we can calculate compatible simultaneous confidence intervals for the individual effects. We derive a small sample size approximation of the distribution of the test statistic via a multivariate t-distribution. As an alternative, we introduce a Wild-bootstrap method. Extensive simulations show that our methods are applicable even when sample sizes are small. Their application is further illustrated within a real data example.

Information technology and data science development stimulate transformation in many fields of scientific knowledge. In recent years, a large number of specialized systems for information and knowledge management have been created in materials science. However, the development and deployment of open adaptive systems for research support in materials science based on the acquisition, storage, and processing of different types of information remains unsolved. We propose MatInf - an extensible, open-source solution for research digitalisation in materials science based on an adaptive, flexible information management system for heterogeneous data sources. MatInf can be easily adapted to any materials science laboratory and is especially useful for collaborative projects between several labs. As an example, we demonstrate its application in high-throughput experimentation., Comment: 7 pages, 5 figures, 10 references

While machine learning (ML) models are increasingly used due to their high predictive power, their use in understanding the data-generating process (DGP) is limited. Understanding the DGP requires insights into feature-target associations, which many ML models cannot directly provide due to their opaque internal mechanisms. Feature importance (FI) methods provide useful insights into the DGP under certain conditions. Since the results of different FI methods have different interpretations, selecting the correct FI method for a concrete use case is crucial and still requires expert knowledge. This paper serves as a comprehensive guide to help understand the different interpretations of global FI methods. Through an extensive review of FI methods and providing new proofs regarding their interpretation, we facilitate a thorough understanding of these methods and formulate concrete recommendations for scientific inference. We conclude by discussing options for FI uncertainty estimation and point to directions for future research aiming at full statistical inference from black-box ML models.

We consider nonreciprocal supercurrent effects in Josephson junctions based on multiband superconductors with a pairing structure that can break time-reversal symmetry. We demonstrate that a nonreciprocal supercurrent can be generally achieved by the cooperation of interband superconducting phase mismatch and interband scattering as well as by multiband phase frustration. The effect of interband impurity scattering indicates that the amplitude and sign of the nonreciprocal supercurrent are sensitive to the interband phase relation. For the case of a three-band superconductor, due to phase frustration, we show that the profile of the supercurrent rectification is marked by a hexagonal pattern of nodal lines with vanishing amplitude. Remarkably, around the nodal lines, the supercurrent rectification amplitude exhibits three-fold structures with an alternating sign. We show that the hexagonal pattern and the three-fold structure in the interband phase space turn out to be dependent on the tunneling amplitude of each band. These findings provide hallmarks of the supercurrent rectification which can be potentially employed to unveil the occurrence of spin-singlet multiband superconductivity with time-reversal symmetry breaking., Comment: 9 pages with 4 figures

In metals and semiconductors, an impurity spin is quantum entangled with and thereby screened by surrounding conduction electrons at low temperatures, called the Kondo screening cloud. Quantum confinement of the Kondo screening cloud in a region, called a Kondo box, with a length smaller than the original cloud extension length strongly deforms the screening cloud and provides a way of controlling the entanglement. Here we realize such a Kondo box and develop an approach to controlling and monitoring the entanglement. It is based on a spin localized in a semiconductor quantum dot, which is screened by conduction electrons along a quasi-one-dimensional channel. The box is formed between the dot and a quantum point contact placed on a channel. As the quantum point contact is tuned to make the confinement stronger, electron conductance through the dot as a function of temperature starts to deviate from the known universal function of the single energy scale, the Kondo temperature. Nevertheless, the entanglement is monitored by the measured conductance according to our theoretical development. The dependence of the monitored entanglement on the confinement strength and temperature implies that the Kondo screening is controlled by tuning the quantum point contact. Namely, the Kondo cloud is deformed by the Kondo box in the region across the original cloud length. Our findings offer a way of manipulating and detecting spatially extended quantum many-body entanglement in solids by electrical means.

We present an iterative optimal control method of quantum systems, aimed at an implementation of a desired operation with optimal fidelity. The update step of the method is based on the linear response of the fidelity to the control operators, and its projection onto the mode functions of the corresponding operator. Our method extends methods such as gradient ascent pulse engineering and variational quantum algorithms, by determining the fidelity gradient in a hyperparameter-free manner, and using it for a multi-parameter update, capitalizing on the multi-mode overlap of the perturbation and the mode functions. This directly reduces the number of dynamical trajectories that need to be evaluated in order to update a set of parameters. We demonstrate this approach, and compare it to the standard GRAPE algorithm, for the example of a quantum gate on two qubits, demonstrating a clear improvement in convergence and optimal fidelity of the generated protocol., Comment: 12 pages, 9 figures

Scene graph generation has emerged as a prominent research field in computer vision, witnessing significant advancements in the recent years. However, despite these strides, precise and thorough definitions for the metrics used to evaluate scene graph generation models are lacking. In this paper, we address this gap in the literature by providing a review and precise definition of commonly used metrics in scene graph generation. Our comprehensive examination clarifies the underlying principles of these metrics and can serve as a reference or introduction to scene graph metrics. Furthermore, to facilitate the usage of these metrics, we introduce a standalone Python package called SGBench that efficiently implements all defined metrics, ensuring their accessibility to the research community. Additionally, we present a scene graph benchmarking web service, that enables researchers to compare scene graph generation methods and increase visibility of new methods in a central place. All of our code can be found at https://lorjul.github.io/sgbench/.

We propose a formalism that captures the algebraic structure of many-body two-level quantum systems, and directly motivates an efficient numerical method. This formalism is based on the binary representation of the enumeration-indices of the elements of the corresponding Lie algebra. The action of arbitrarily large elements of that algebra reduces to a few bit-wise exclusive-or operations. This formalism naturally produces sparse representations of many-body density operators, the size of which we control through a dynamic truncation method. We demonstrate how this formalism applies to real-time evolution, dissipative Lindblad action, imaginary-time evolution, and projective measurement processes. We find that this approach to calculating quantum dynamics scales close to linearly with the number of non-zero components in the density operator. We refer to this exclusive-or represented quantum algebra as ORQA. As a proof of concept, we provide a numerical demonstration of this formalism by simulating quantum annealing processes for the maximum independent set problem for up to 22 two-level systems.

B. A. Barnes introduced so-called Fredholm elements in a semiprime ring whose definition is inspired by Atkinson's theorem. Here the socle of a semiprime ring generalizes the ideal of finite-rank operators on a Banach space. In this paper, we aim to see that the algebraic concept of the length of a module is strongly related to that of Fredholm elements. This motivates another generalization of Fredholm elements by requiring for an element $a\in\mathcal{A}$ that the $\mathcal{A}$-modules of the form $\mathcal{A}/\mathcal{A} a$ and $\mathcal{A}/a\mathcal{A}$ are of finite length. We are particularly interested in sufficient conditions for our generalized Fredholm elements to be Fredholm. In a unital C$^*$-algebra $\mathcal{A}$ we shall even see that an element $a\in\mathcal{A}$ is Fredholm if and only if the $\mathcal{A}$-modules $\mathcal{A}/\mathcal{A} a$ and $\mathcal{A}/a\mathcal{A}$ both have finite length.

The interactive segmentation task consists in the creation of object segmentation masks based on user interactions. The most common way to guide a model towards producing a correct segmentation consists in clicks on the object and background. The recently published Segment Anything Model (SAM) supports a generalized version of the interactive segmentation problem and has been trained on an object segmentation dataset which contains 1.1B masks. Though being trained extensively and with the explicit purpose of serving as a foundation model, we show significant limitations of SAM when being applied for interactive segmentation on novel domains or object types. On the used datasets, SAM displays a failure rate $\text{FR}_{30}@90$ of up to $72.6 \%$. Since we still want such foundation models to be immediately applicable, we present a framework that can adapt SAM during immediate usage. For this we will leverage the user interactions and masks, which are constructed during the interactive segmentation process. We use this information to generate pseudo-labels, which we use to compute a loss function and optimize a part of the SAM model. The presented method causes a relative reduction of up to $48.1 \%$ in the $\text{FR}_{20}@85$ and $46.6 \%$ in the $\text{FR}_{30}@90$ metrics., Comment: 11 pages, 2 figures, 4 tables

Searches for small exoplanets around solar-type stars are limited by stellar physical variability. While chromospheric variability is well studied, observing, modeling. and understanding the much smaller fluctuations in photospheric spectral line strengths, shapes, and shifts is challenging. Extreme precision radial-velocity spectrometers now enable extreme precision stellar spectroscopy and time series of the Sun seen as a star permit monitoring of its photospheric variability. Fluctuations in their line strengths may well correlate with radial-velocity excursions and identify observable proxies for their monitoring. From three years of HARPS-N observations of the Sun-as-a-star, one thousand low-noise spectra are selected, and line absorption measured in Fe I, Fe II, Mg I, Mn I, H-alpha, H-beta, H-gamma, Na I, and the G-band. Their variations and likely atmospheric origins are examined, also with respect to simultaneously measured chromospheric emission and apparent radial velocity. Systematic line-strength variability is seen, largely shadowing the solar-cycle evolution of Ca II H & K emission, but with smaller amounts, typically on a sub-percent level. Among iron lines, greatest amplitudes are for Fe II in the blue, while the trends change sign among differently strong lines in the green Mg I triplet and between Balmer lines. Variations in the G-band core are greater than of the full G-band, in line with theoretical predictions. No variation is detected in the semi-forbidden Mg I 457.1 nm. Hyperfine split Mn I behaves largely similar to Fe I. For lines at longer wavelengths, telluric absorption limits the achievable precision. Microvariability in the solar photospheric spectrum thus displays systematic signatures among various features. These measure something different than the classical Ca II H & K index while still reflecting a strong influence from magnetic regions., Comment: 20 pages, 22 figures, accepted for publication in Astronomy & Astrophysics

One key task in environmental science is to map environmental variables continuously in space or even in space and time. Machine learning algorithms are frequently used to learn from local field observations to make spatial predictions by estimating the value of the variable of interest in places where it has not been measured. However, the application of machine learning strategies for spatial mapping involves additional challenges compared to "non-spatial" prediction tasks that often originate from spatial autocorrelation and from training data that are not independent and identically distributed. In the past few years, we developed a number of methods to support the application of machine learning for spatial data which involves the development of suitable cross-validation strategies for performance assessment and model selection, spatial feature selection, and methods to assess the area of applicability of the trained models. The intention of the CAST package is to support the application of machine learning strategies for predictive mapping by implementing such methods and making them available for easy integration into modelling workflows. Here we introduce the CAST package and its core functionalities. At the case study of mapping plant species richness, we will go through the different steps of the modelling workflow and show how CAST can be used to support more reliable spatial predictions., Comment: 16 pages,9 figures

We report on the modeling and experimental characterization of Si CMOS detectors of terahertz radiation based on antenna-coupled field-effect transistors (TeraFETs). The detectors are manufactured using TSMC's 65-nm technology. We apply two models -- the TSMC RF foundry model and our own ADS-HDM -- to simulate the Si CMOS TeraFET performance and compare their predictions with respective experimental data. Both models are implemented in the commercial circuit simulation software Keysight Advanced Design System (ADS). We find that the compact model TSMC RF is capable to predict the detector responsivity and its dependence on frequency and gate voltage with good accuracy up to the highest frequency of 1.2 THz covered in this study. This frequency is well beyond the tool's intended operation range for 5G communications and 110-GHz millimeter wave applications. We demonstrate that our self-developed physics-based ADS-HDM tool, which relies on an extended one-dimensional hydrodynamic transport model and can be adapted readily to other material technologies, has high predictive qualities comparable to those of the foundry model. We use the ADS-HDM to discuss the contribution of diffusive and plasmonic effects to the THz response of Si CMOS TeraFETs, finding that these effects, while becoming more significant with rising frequency, are never dominant. Finally, we estimate that the electrical NEP (perfect power coupling conditions) is on the order of 5 pW/$\sqrt{\rm{Hz}}$ at room-temperature.

We investigate the dynamics of an ion moving through a homogeneous Bose-Einstein condensate (BEC) after an initial momentum is imparted. For this, we derive a master equation in the weak-coupling limit and Lamb-Dicke approximation for the reduced density matrix of the ion. We study the time evolution of the ion's kinetic energy and observe that its expectation value, identified as the ion temperature $T_\mathrm{ion}$, is reduced by several orders of magnitude in a time on the order of microseconds for a condensate density in the experimentally relevant range between $10^{13}\,\mathrm{cm}^{-3}$ and $10^{14}\,\mathrm{cm}^{-3}$. We characterize this behavior by defining the duration at half maximum as the time required by $T_\mathrm{ion}$ to reach half of its initial value, and study its dependence on the system parameters. Similarly, we find that the expectation value of the ion's momentum operator is reduced by nine orders of magnitude on the same timescale, making the ion's position converge to a final value. Based on these results, we conclude that the interaction with the bosonic bath allows for cooling and pinning of the ion by decreasing the expectation value of its kinetic energy and velocity, which constitutes a result of direct relevance for current atom-ion experiments., Comment: 16 pages, 11 figures

The physical processes responsible for shaping how galaxies form and quench over time leave imprints on both the spatial (galaxy morphology) and temporal (star formation history; SFH) tracers that we use to study galaxies. While the morphology-SFR connection is well studied, the correlation with past star formation activity is not as well understood. To quantify this we present Katachi, an interpretable convolutional neural network (CNN) framework that learns the connection between the factors regulating star formation in galaxies on different spatial and temporal scales. Katachi is trained on 9904 galaxies at 0.02$<$z$<$0.1 in the SDSS-IV MaNGA DR17 sample to predict stellar mass (M$_*$; RMSE 0.22 dex), current star formation rate (SFR; RMSE 0.31 dex) and half-mass time (t$_{50}$; RMSE 0.23 dex). This information allows us to reconstruct non-parametric SFHs for each galaxy from \textit{gri} imaging alone. To quantify the morphological features informing the SFH predictions we use SHAP (SHapley Additive exPlanations). We recover the expected trends of M$_*$ governed by the growth of galaxy bulges, and SFR correlating with spiral arms and other star-forming regions. We also find the SHAP maps of D4000 are more complex than those of M$_*$ and SFR, and that morphology is correlated with t$_{50}$ even at fixed mass and SFR. Katachi serves as a scalable public framework to predict galaxy properties from large imaging surveys including Rubin, Roman, and Euclid, with large datasets of high SNR imaging across limited photometric bands., Comment: 35 pages, 25 figures. Accepted for publication in ApJ, comments are welcomed

One of the key shortcomings in current text-to-image (T2I) models is their inability to consistently generate images which faithfully follow the spatial relationships specified in the text prompt. In this paper, we offer a comprehensive investigation of this limitation, while also developing datasets and methods that support algorithmic solutions to improve spatial reasoning in T2I models. We find that spatial relationships are under-represented in the image descriptions found in current vision-language datasets. To alleviate this data bottleneck, we create SPRIGHT, the first spatially focused, large-scale dataset, by re-captioning 6 million images from 4 widely used vision datasets and through a 3-fold evaluation and analysis pipeline, show that SPRIGHT improves the proportion of spatial relationships in existing datasets. We show the efficacy of SPRIGHT data by showing that using only $\sim$0.25% of SPRIGHT results in a 22% improvement in generating spatially accurate images while also improving FID and CMMD scores. We also find that training on images containing a larger number of objects leads to substantial improvements in spatial consistency, including state-of-the-art results on T2I-CompBench with a spatial score of 0.2133, by fine-tuning on <500 images. Through a set of controlled experiments and ablations, we document additional findings that could support future work that seeks to understand factors that affect spatial consistency in text-to-image models., Comment: Accepted to ECCV 2024. Project Page : https://spright-t2i.github.io/

Recent work shows that path gradient estimators for normalizing flows have lower variance compared to standard estimators for variational inference, resulting in improved training. However, they are often prohibitively more expensive from a computational point of view and cannot be applied to maximum likelihood training in a scalable manner, which severely hinders their widespread adoption. In this work, we overcome these crucial limitations. Specifically, we propose a fast path gradient estimator which improves computational efficiency significantly and works for all normalizing flow architectures of practical relevance. We then show that this estimator can also be applied to maximum likelihood training for which it has a regularizing effect as it can take the form of a given target energy function into account. We empirically establish its superior performance and reduced variance for several natural sciences applications.

Complete quantum control of a stationary quantum bit embedded in a quantum emitter is crucial for photonic quantum information technologies. Recently, the orbital degree of freedom in optically active quantum dots has emerged as a promising candidate. However, the essential ability to perform arbitrary rotations on orbital qubits remains elusive. Here, we demonstrate arbitrary rotation of a hole orbital qubit with direct phase control using picosecond optical pulses. This is achieved by successfully inducing stimulated Raman transitions within $\Lambda$ systems coupled via radiative Auger processes. The new capability enables direct control of polar and azimuth angles of the Bloch vector without requiring timed precession. Our results establish orbital states in solid-state quantum emitters as a viable resource for applications in high-speed quantum information processing., Comment: Manuscript with 7 pages and 4 figures plus Supplementary Material comprising 9 pages and 10 figures

Large vision language models, such as CLIP, demonstrate impressive robustness to spurious features than single-modal models trained on ImageNet. However, existing test datasets are typically curated based on ImageNet-trained models, which aim to capture the spurious features inherited in ImageNet. Benchmarking CLIP models based on the ImageNet-oriented spurious features may not be sufficient to reflect the extent to which CLIP models are robust to spurious correlations within CLIP training data, e.g., LAION. To this end, we craft a new challenging dataset named CounterAnimal designed to reveal the reliance of CLIP models on realistic spurious features. Specifically, we split animal photos into groups according to the backgrounds, and then identify a pair of groups for each class where a CLIP model shows high-performance drops across the two groups. Our evaluations show that the spurious features captured by CounterAnimal are generically learned by CLIP models with different backbones and pre-train data, yet have limited influence for ImageNet models. We provide theoretical insights that the CLIP objective cannot offer additional robustness. Furthermore, we also re-evaluate strategies such as scaling up parameters and high-quality pre-trained data. We find that they still help mitigate the spurious features, providing a promising path for future developments., Comment: NeurIPS 2024; Qizhou Wang, Yong Lin, and Yongqiang Chen contributed equally; Project page: https://counteranimal.github.io

Quantum circuit synthesis describes the process of converting arbitrary unitary operations into a gate sequence of a fixed universal gate set, usually defined by the operations native to a given hardware platform. Most current synthesis algorithms are designed to synthesize towards a set of single qubit rotations and an additional entangling two qubit gate, such as CX, CZ, or the Molmer Sorensen gate. However, with the emergence of neutral atom based hardware and their native support for gates with more than two qubits, synthesis approaches tailored to these new gate sets become necessary. In this work, we present an approach to synthesize multi controlled phase gates using ZX calculus. By representing quantum circuits as graph like ZX diagrams, one can utilize the distinct graph structure of diagonal gates to identify multi controlled phase gates inherently present in some quantum circuits even if none were explicitly defined in the original circuit. We evaluate the approach on a wide range of benchmark circuits and compare them to the standard Qiskit synthesis regarding its circuit execution time for neutral atom based hardware with native support of multi controlled gates. Our results show possible advantages for current state of the art hardware and represent the first exact synthesis algorithm supporting arbitrary sized multi controlled phase gates., Comment: In Proceedings QPL 2024, arXiv:2408.05113

Scaling laws are useful guides for derisking expensive training runs, as they predict performance of large models using cheaper, small-scale experiments. However, there remain gaps between current scaling studies and how language models are ultimately trained and evaluated. For instance, scaling is usually studied in the compute-optimal training regime (i.e., "Chinchilla optimal" regime). In contrast, models are often over-trained to reduce inference costs. Moreover, scaling laws mostly predict loss on next-token prediction, but models are usually compared on downstream task performance. To address both shortcomings, we create a testbed of 104 models with 0.011B to 6.9B parameters trained with various numbers of tokens on three data distributions. First, we fit scaling laws that extrapolate in both the amount of over-training and the number of model parameters. This enables us to predict the validation loss of a 1.4B parameter, 900B token run (i.e., 32$\times$ over-trained) and a 6.9B parameter, 138B token run (i.e., a compute-optimal run)$\unicode{x2014}$each from experiments that take 300$\times$ less compute. Second, we relate the perplexity of a language model to its downstream task performance by proposing a power law. We use this law to predict top-1 error averaged over downstream tasks for the two aforementioned models, using experiments that take 20$\times$ less compute. Our experiments are available at https://github.com/mlfoundations/scaling.

The quantum cellular automata (QCA) effect is a transition in which multiple electron move coordinately by Coulomb interactions and observed in multiple quantum dots. This effect will be useful for realizing and improving quantum cellular automata and information transfer using multiple electron transfer. In this paper, we investigate the real-time dynamics of the QCA charge transitions in a triple quantum dot by using fast charge-state readout realized by rf reflectometry. We observe real-time charge transitions and analyze the tunneling rate comparing with the first-order tunneling processes. We also measure the gate voltage dependence of the QCA transition and show that it can be controlled by the voltage., Comment: 11pages, 4 figures

We present optimal control strategies for the DC transport across a Josephson junction. Specifically, we consider a junction in which the Josephson coupling is driven parametrically, with either a bichromatic or a trichromatic driving protocol, and optimize the prefactor of the 1/$\omega$ divergence of the imaginary part of the conductivity. We demonstrate that for an optimal bichromatic protocol an enhancement of 70 can be reached, and for an optimal trichromatic protocol an enhancement of 135. This is motivated by pump-probe experiments that have demonstrated light-enhanced superconductivity along the c-axis of underdoped YBCO, where the junction serves as a minimal model for the c-axis coupling of superconducting layers. Therefore, the significant enhancement of superconductivity that we show for multi-frequency protocols demonstrates that the advancement of pump-probe technology towards these strategies is highly desirable.

The local atomic structure of NaF-ZrF$_4$ (53-47 mol%) molten system and its evolution with temperature are examined with x-ray scattering measurements and compared with $ab-initio$ and Neural Network-based molecular dynamics (NNMD) simulations in the temperature range 515-700 {\deg}C. The machine-learning enhanced NNMD calculations offer improved efficiency while maintaining accuracy at higher distances compared to ab-initio calculations. Looking at the evolution of the Pair Distribution Function with increasing temperature, a fundamental change in the liquid structure within the selected temperature range, accompanied by a slight decrease in overall correlation is revealed. NNMD calculations indicate the co-existence of three different fluorozirconate complexes: [ZrF$_6$]$^{2-}$, [ZrF$_7$]$^{3-}$, and [ZrF$_8$]$^{4-}$, with a temperature-dependent shift in the dominant coordination state towards a 6-coordinated Zr ion at 700{\deg}C. The study also highlights the metastability of different coordination structures, with frequent interconversions between 6 and 7 coordinate states for the fluorozirconate complex from 525 {\deg}C to 700 {\deg}C. Analysis of the Zr-F-Zr angular distribution function reveals the presence of both $"$edge-sharing$"$ and $"$corner-sharing$"$ fluorozirconate complexes with specific bond angles and distances in accord with previous studies, while the next-nearest neighbor cation-cation correlations demonstrate a clear preference for unlike cations as nearest-neighbor pairs, emphasizing non-random arrangement. These findings contribute to a comprehensive understanding of the complex local structure of the molten salt, providing insights into temperature-dependent preferences and correlations within the molten system., Comment: 26 pages, 15 figures, 3 tables

Multiple access techniques are cornerstones of wireless communications. Their performance depends on the channel properties, which can be improved by reconfigurable intelligent surfaces (RISs). In this work, we jointly optimize MA precoding at the base station (BS) and RIS configuration. We tackle difficulties of mutual coupling between RIS elements, scalability to more than 1000 RIS elements, and channel estimation. We first derive an RIS-assisted channel model considering mutual coupling, then propose an unsupervised machine learning (ML) approach to optimize the RIS. In particular, we design a dedicated neural network (NN) architecture RISnet with good scalability and desired symmetry. Moreover, we combine ML-enabled RIS configuration and analytical precoding at BS since there exist analytical precoding schemes. Furthermore, we propose another variant of RISnet, which requires the channel state information (CSI) of a small portion of RIS elements (in this work, 16 out of 1296 elements) if the channel comprises a few specular propagation paths. More generally, this work is an early contribution to combine ML technique and domain knowledge in communication for NN architecture design. Compared to generic ML, the problem-specific ML can achieve higher performance, lower complexity and symmetry., Comment: 13 pages, 16 figures