Search

Although by now the ensemble-based probabilistic forecasting is the most advanced approach to weather prediction, ensemble forecasts still might suffer from lack of calibration and/or display systematic bias, thus require some post-processing to improve their forecast skill. Here we focus on visibility, which quantity plays a crucial role e.g. in aviation and road safety or in ship navigation, and propose a parametric model where the predictive distribution is a mixture of a gamma and a truncated normal distribution, both right censored at the maximal reported visibility value. The new model is evaluated in two case studies based on visibility ensemble forecasts of the European Centre for Medium-Range Weather Forecasts covering two distinct domains in Central and Western Europe and two different time periods. The results of the case studies indicate that post-processed forecasts are substantially superior to the raw ensemble; moreover, the proposed mixture model consistently outperforms the Bayesian model averaging approach used as reference post-processing technique., Comment: 26 pages, 14 figures, 3 tables

Unsupervised Semantic Segmentation (USS) involves segmenting images without relying on predefined labels, aiming to alleviate the burden of extensive human labeling. Existing methods utilize features generated by self-supervised models and specific priors for clustering. However, their clustering objectives are not involved in the optimization of the features during training. Additionally, due to the lack of clear class definitions in USS, the resulting segments may not align well with the clustering objective. In this paper, we introduce a novel approach called Optimally Matched Hierarchy (OMH) to simultaneously address the above issues. The core of our method lies in imposing structured sparsity on the feature space, which allows the features to encode information with different levels of granularity. The structure of this sparsity stems from our hierarchy (OMH). To achieve this, we learn a soft but sparse hierarchy among parallel clusters through Optimal Transport. Our OMH yields better unsupervised segmentation performance compared to existing USS methods. Our extensive experiments demonstrate the benefits of OMH when utilizing our differentiable paradigm. We will make our code publicly available., Comment: 11 pages

Simulating ultra-high-granularity detector responses in Particle Physics represents a critical yet computationally demanding task. This thesis aims to overcome this challenge for the Pixel Vertex Detector (PXD) at the Belle II experiment, which features over 7.5M pixel channels-the highest spatial resolution detector simulation dataset ever analysed with generative models. This thesis starts off by a comprehensive and taxonomic review on generative models for simulating detector signatures. Then, it presents the Intra-Event Aware Generative Adversarial Network (IEA-GAN), a new geometry-aware generative model that introduces a relational attentive reasoning and Self-Supervised Learning to approximate an "event" in the detector. This study underscores the importance of intra-event correlation for downstream physics analyses. Building upon this, the work drifts towards a more generic approach and presents YonedaVAE, a Category Theory-inspired generative model that tackles the open problem of Out-of-Distribution (OOD) simulation. YonedaVAE introduces a learnable Yoneda embedding to capture the entirety of an event based on its sensor relationships, formulating a Category theoretical language for intra-event relational reasoning. This is complemented by introducing a Self-Supervised learnable prior for VAEs and an Adaptive Top-q sampling mechanism, enabling the model to sample point clouds with variable intra-category cardinality in a zero-shot manner. Variable Intra-event cardinality has not been approached before and is vital for simulating irregular detector geometries. Trained on an early experiment data, YonedaVAE can reach a reasonable OOD simulation precision of a later experiment with almost double luminosity. This study introduces, for the first time, the results of using deep generative models for ultra-high granularity detector simulation in Particle Physics., Comment: PhD thesis, 234 pages

We present results of a continuation of our Transiting Exoplanet Survey Satellite (TESS) search for short-period pulsations in compact stellar objects observed during Years 2 and 4 of the TESS mission that targeted the northern ecliptic hemisphere. For many of the targets, we exploit unpublished spectroscopic data to confirm or determine the object's spectral classification. From the TESS photometry, we identify 50 short-period hot-subdwarf pulsators, including 35 sdB and 15 sdOB stars. The sample contains 26 pulsators not known before the TESS mission. Nine stars show signals at both low and high frequencies, and are therefore ``hybrid'' pulsators. For each pulsator, we report the list of prewhitened frequencies and we show amplitude spectra calculated from the TESS data. We attempt to identify possible multiplets caused by stellar rotation, and we report five candidates with rotation periods between 11 and 46d. Having the search for p-mode pulsating hot subdwarfs in TESS Sectors 1 - 60 done, we discuss the completeness of the study, as well as instability strip and the evolutionary status of the stars we found. We also compare the distribution of pulsation periods as a function of effective temperature and surface gravity with theoretical predictions., Comment: accepted

Superconductor-semiconductor hybrid devices, involving quantum dots interfaced with floating and/or grounded superconductors, have reached a level of complexity which calls for the development of versatile and numerically efficient modelling tools. Here, we propose an extension of the surrogate model solver for sub-gap states [Phys. Rev. B 108, L220506 (2023)], which is able to handle floating superconducting islands with finite charging energy. Upon eliminating all finite-size effects of the computationally demanding Richardson model approach, we achieve a more efficient way of calculating the sub-gap spectra and related observables without compromising their accuracy. We provide a number of benchmarks between the two approaches and showcase the versatility of the extended surrogate model solver by studying the stability of spin-triplet ground states in various tunable devices. The methods introduced here set the stage for reliable microscopic simulations of complex superconducting quantum circuits across all their relevant parameter regimes., Comment: 12 pages, 12 figures

Magnetocaloric performance of the RE$_{5}$Pd$_2$In$_4$ (RE = Tb-Tm) rare earth compounds has been investigated using measurements of magnetization in the function of temperature and applied magnetic field. The maximum magnetic entropy change ($-\Delta S_{M}^{max}$) at magnetic flux density change ($\Delta \mu_0 H$) of 0-9~T has been determined to be 3.3~J$\cdot$kg$^{-1}\cdot$K$^{-1}$ at 62~K for Tb$_{5}$Pd$_2$In$_4$, 7.0~J$\cdot$kg$^{-1}\cdot$K$^{-1}$ at 22~K for Dy$_{5}$Pd$_2$In$_4$, 12.6~J$\cdot$kg$^{-1}\cdot$K$^{-1}$ at 22~K for Ho$_{5}$Pd$_2$In$_4$, 12.1~J$\cdot$kg$^{-1}\cdot$K$^{-1}$ at 17~K for Er$_{5}$Pd$_2$In$_4$ and 11.9~J$\cdot$kg$^{-1}\cdot$K$^{-1}$ at 9.0~K for Tm$_{5}$Pd$_2$In$_4$. The temperature averaged entropy change (TEC) with 3~K span equals 3.2, 7.0, 12.6, 12.2 and 11.8~J$\cdot$kg$^{-1}\cdot$K$^{-1}$ for RE = Tb-Tm, respectively. The relative cooling power (RCP) and refrigerant capacity (RC) are equal to respectively 258 and 215~J$\cdot$kg$^{-1}$ in Tb$_{5}$Pd$_2$In$_4$, 498 and 325~J$\cdot$kg$^{-1}$ in Dy$_{5}$Pd$_2$In$_4$, 489 and 403~J$\cdot$kg$^{-1}$ in Ho$_{5}$Pd$_2$In$_4$, 403 and 314~J$\cdot$kg$^{-1}$ in Er$_{5}$Pd$_2$In$_4$ and 234 and 184~J$\cdot$kg$^{-1}$ in Tm$_{5}$Pd$_2$In$_4$. The magnetocaloric properties of RE$_{5}$Pd$_2$In$_4$ are comparable to those of other known magnetocaloric materials, which show good magnetocaloric performance at low temperatures. Among RE$_{5}$Pd$_2$In$_4$, the highest values of parameters characterizing the magnetocaloric effect are found for RE = Ho and Er. Furthermore, for fixed RE element, the RE$_{5}$Pd$_2$In$_4$ compound displays the highest RCP and RC values when compared to those of its isostructural RE$_{5}$T$_2$In$_4$ (T = Ni, Pt) analogues., Comment: 9 pages, 3 figures. arXiv admin note: text overlap with arXiv:2212.07171

By computing a feedback control via the linear quadratic regulator (LQR) approach and simulating a non-linear non-autonomous closed-loop system using this feedback, we combine two numerically challenging tasks. For the first task, the computation of the feedback control, we use the non-autonomous generalized differential Riccati equation (DRE), whose solution determines the time-varying feedback gain matrix. Regarding the second task, we want to be able to simulate non-linear closed-loop systems for which it is known that the regulator is only valid for sufficiently small perturbations. Thus, one easily runs into numerical issues in the integrators when the closed-loop control varies greatly. For these systems, e.g., the A-stable implicit Euler methods fails.\newline On the one hand, we implement non-autonomous versions of splitting schemes and BDF methods for the solution of our non-autonomous DREs. These are well-established DRE solvers in the autonomous case. On the other hand, to tackle the numerical issues in the simulation of the non-linear closed-loop system, we apply a fractional-step-theta scheme with time-adaptivity tuned specifically to this kind of challenge. That is, we additionally base the time-adaptivity on the activity of the control. We compare this approach to the more classical error-based time-adaptivity.\newline We describe techniques to make these two tasks computable in a reasonable amount of time and are able to simulate closed-loop systems with strongly varying controls, while avoiding numerical issues. Our time-adaptivity approach requires fewer time steps than the error-based alternative and is more reliable.

Although Machine Learning model building has become increasingly accessible due to a plethora of tools, libraries and algorithms being available freely, easy operationalization of these models is still a problem. It requires considerable expertise in data engineering, software development, cloud and DevOps. It also requires planning, agreement, and vision of how the model is going to be used by the business applications once it is in production, how it is going to be continuously trained on fresh incoming data, and how and when a newer model would replace an existing model. This leads to developers and data scientists working in silos and making suboptimal decisions. It also leads to wasted time and effort. We introduce the Acumos AI platform we developed and we demonstrate some unique novel capabilities that the Acumos model runner possesses, that can help solve the above problems. We introduce a new sustainable concept in the field of AI/ML operations - called Reusable MLOps - where we reuse the existing deployment and infrastructure to serve new models by hot-swapping them without tearing down the infrastructure or the microservice, thus achieving reusable deployment and operations for AI/ML models while still having continuously trained models in production.

The ice surface is known for presenting a very small kinetic friction coefficient, but the origin of this property remains highly controversial to date. In this work, we revisit recent computer simulations of ice sliding on atomically smooth substrates, using newly calculated bulk viscosities for the TIP4P/Ice water model. The results show that spontaneously formed premelting films in static conditions exhibit an effective viscosity which is about twice the bulk viscosity. However, upon approaching sliding speeds in the order of m/s, the shear rate becomes very large, and the viscosities decrease by several orders of magnitude. This shows that premelting films can act as an efficient lubrication layer despite their small thickness, and illustrates an interesting interplay between confinement enhanced viscosities, and shear thinning. Our results suggest that the strongly thinned viscosities that operate under the high speed skating regime could largely reduce the amount of frictional heating., Comment: 5 pages, 1 figure

This paper describes the results of the IEEE BigData 2023 Keystroke Verification Challenge (KVC), that considers the biometric verification performance of Keystroke Dynamics (KD), captured as tweet-long sequences of variable transcript text from over 185,000 subjects. The data are obtained from two of the largest public databases of KD up to date, the Aalto Desktop and Mobile Keystroke Databases, guaranteeing a minimum amount of data per subject, age and gender annotations, absence of corrupted data, and avoiding excessively unbalanced subject distributions with respect to the considered demographic attributes. Several neural architectures were proposed by the participants, leading to global Equal Error Rates (EERs) as low as 3.33% and 3.61% achieved by the best team respectively in the desktop and mobile scenario, outperforming the current state of the art biometric verification performance for KD. Hosted on CodaLab, the KVC will be made ongoing to represent a useful tool for the research community to compare different approaches under the same experimental conditions and to deepen the knowledge of the field., Comment: 9 pages, 10 pages, 2 figures. arXiv admin note: text overlap with arXiv:2311.06000

Active soft bodies can affect their shape through an internal actuation mechanism that induces a deformation. Similar to recent work, this paper utilizes a differentiable, quasi-static, and physics-based simulation layer to optimize for actuation signals parameterized by neural networks. Our key contribution is a general and implicit formulation to control active soft bodies by defining a function that enables a continuous mapping from a spatial point in the material space to the actuation value. This property allows us to capture the signal's dominant frequencies, making the method discretization agnostic and widely applicable. We extend our implicit model to mandible kinematics for the particular case of facial animation and show that we can reliably reproduce facial expressions captured with high-quality capture systems. We apply the method to volumetric soft bodies, human poses, and facial expressions, demonstrating artist-friendly properties, such as simple control over the latent space and resolution invariance at test time., Comment: Accepted to SIGGRAPH 2022. Project page: https://studios.disneyresearch.com/2022/07/24/implicit-neural-representation-for-physics-driven-actuated-soft-bodies/ Video: https://www.youtube.com/watch?v=9EERe_CTazk

Since the start of the operational use of ensemble prediction systems, ensemble-based probabilistic forecasting has become the most advanced approach in weather prediction. However, despite the persistent development of the last three decades, ensemble forecasts still often suffer from the lack of calibration and might exhibit systematic bias, which calls for some form of statistical post-processing. Nowadays, one can choose from a large variety of post-processing approaches, where parametric methods provide full predictive distributions of the investigated weather quantity. Parameter estimation in these models is based on training data consisting of past forecast-observation pairs, thus post-processed forecasts are usually available only at those locations where training data are accessible. We propose a general clustering-based interpolation technique of extending calibrated predictive distributions from observation stations to any location in the ensemble domain where there are ensemble forecasts at hand. Focusing on the ensemble model output statistics (EMOS) post-processing technique, in a case study based on wind speed ensemble forecasts of the European Centre for Medium-Range Weather Forecasts, we demonstrate the predictive performance of various versions of the suggested method and show its superiority over the regionally estimated and interpolated EMOS models and the raw ensemble forecasts as well., Comment: 19 pages, 6 figures

State-of-the-art models can perform well in controlled environments, but they often struggle when presented with out-of-distribution (OOD) examples, making OOD detection a critical component of NLP systems. In this paper, we focus on highlighting the limitations of existing approaches to OOD detection in NLP. Specifically, we evaluated eight OOD detection methods that are easily integrable into existing NLP systems and require no additional OOD data or model modifications. One of our contributions is providing a well-structured research environment that allows for full reproducibility of the results. Additionally, our analysis shows that existing OOD detection methods for NLP tasks are not yet sufficiently sensitive to capture all samples characterized by various types of distributional shifts. Particularly challenging testing scenarios arise in cases of background shift and randomly shuffled word order within in domain texts. This highlights the need for future work to develop more effective OOD detection approaches for the NLP problems, and our work provides a well-defined foundation for further research in this area., Comment: 11 pages, 3 figures, Association for Computational Linguistics

While in traditional mathematics teaching methods, it is possible to prove concepts while teaching geometry and to reach generalizations that require more complex definitions by using concept findings, while in today's geometry teaching, learning that activates the student's discovery process gains importance. In this context, the aim of this study is to examine ten different studies on the factors affecting students' learning of basic geometric concepts. In this study, document analysis method was used. The main purpose of document review is not to reach new information, but to reveal the current situation in light of different studies. During the data collection process of the research, ten different articles on the factors affecting learning basic geometric concepts in geometry teaching were examined by scanning Google Scholar. For the research, a keyword in the form of factors affecting geometry teaching was entered into the search engine at the scanning stage and ten different articles that were allowed to access full texts were examined. For the analysis of the data, 10 different studies made by the researchers were analyzed as a result of the document review. During the document analysis, descriptive analysis was carried out by looking at the sample, sample selection method, data analysis method, method and findings of the studies examined by the researcher, and the findings were presented in tables. As a result of the research, it was concluded that constructivist learning theories and methods are effective in teaching basic geometric concepts. [This paper was published in: "EJER Congress 2023 International Eurasian Educational Research Congress Conference Proceedings," Ani Publishing, 2023, pp. 645-657.]