Search

Spherical k-means is a widely used clustering algorithm for sparse and high-dimensional data such as document vectors. While several improvements and accelerations have been introduced for the original k-means algorithm, not all easily translate to the spherical variant: Many acceleration techniques, such as the algorithms of Elkan and Hamerly, rely on the triangle inequality of Euclidean distances. However, spherical k-means uses Cosine similarities instead of distances for computational efficiency. In this paper, we incorporate the Elkan and Hamerly accelerations to the spherical k-means algorithm working directly with the Cosines instead of Euclidean distances to obtain a substantial speedup and evaluate these spherical accelerations on real data.

FU Orionis-type objects (FUors) are low-mass pre-main sequence stars undergoing a temporary, but significant increase of mass accretion rate from the circumstellar disk onto the protostar. It is not yet clear what triggers the accretion bursts and whether the disks of FUors are in any way different from disks of non-bursting young stellar objects. Motivated by this, we conducted a 1.3 mm continuum survey of ten FUors and FUor-like objects with ALMA, using both the 7 m array and the 12 m array in two different configurations to recover emission at the widest possible range of spatial scales. We detected all targeted sources and several nearby objects as well. To constrain the disk structure, we fit the data with models of increasing complexity from 2D Gaussian to radiative transfer, enabling comparison with other samples modeled in a similar way. The radiative transfer modeling gives disk masses that are significantly larger than what is obtained from the measured millimeter fluxes assuming optically thin emission, suggesting that the FUor disks are optically thick at this wavelength. In comparison with samples of regular Class II and Class I objects, the disks of FUors are typically a factor of 2.9-4.4 more massive and a factor of 1.5-4.7 smaller in size. A significant fraction of them (65-70%) may be gravitationally unstable., Comment: 34 pages, 8 tables, 29 figures, accepted for publication in the Astrophysical Journal Supplement Series

Smart electricity meters typically upload readings a few times a day. Utility providers aim to increase the upload frequency in order to access consumption information in near real time, but the legacy compressors fail to provide sufficient savings on the low-bandwidth, high-cost data connection. We propose a new compression method and data format for DLMS smart meter readings, which is significantly better with frequent uploads and enable reporting every reading in near real time with the same or lower data sizes than the currently available compressors in the DLMS protocol., Comment: 6 pages, 7 figures, IEEE conference format, submitted to Globecom'21

The radio spectra of main-sequence stars remain largely unconstrained due to the lack of observational data to inform stellar atmosphere models. As such, the dominant emission mechanisms at long wavelengths, how they vary with spectral type, and how much they contribute to the expected brightness at a given radio wavelength are still relatively unknown for most spectral types. We present radio continuum observations of Altair, a rapidly rotating A-type star. We observed Altair with NOEMA in 2018 and 2019 at 1.34 mm, 2.09 mm, and 3.22 mm and with the VLA in 2019 at 6.7 mm and 9.1 mm. In the radio spectra, we see a brightness temperature minimum at millimeter wavelengths followed by a steep rise to temperatures larger than the optical photosphere, behavior that is unexpected for A-type stars. We use these data to produce the first sub-millimeter to centimeter spectrum of a rapidly rotating A-type star informed by observations. We generated both PHOENIX and KINICH-PAKAL model atmospheres and determine the KINICH-PAKAL model better reproduces Altair's radio spectrum. The synthetic spectrum shows a millimeter brightness temperature minimum followed by significant emission over that of the photosphere at centimeter wavelengths. Together, these data and models show how the radio spectrum of an A-type star can reveal the presence of a chromosphere, likely induced by rapid rotation, and that a Rayleigh Jean's extrapolation of the stellar photosphere is not an adequate representation of a star's radio spectrum., Comment: 11 pages, 3 figures, accepted to ApJL

Currently, a growing number of mature natural language processing applications make people's life more convenient. Such applications are built by source code - the language in software engineering. However, the applications for understanding source code language to ease the software engineering process are under-researched. Simultaneously, the transformer model, especially its combination with transfer learning, has been proven to be a powerful technique for natural language processing tasks. These breakthroughs point out a promising direction for process source code and crack software engineering tasks. This paper describes CodeTrans - an encoder-decoder transformer model for tasks in the software engineering domain, that explores the effectiveness of encoder-decoder transformer models for six software engineering tasks, including thirteen sub-tasks. Moreover, we have investigated the effect of different training strategies, including single-task learning, transfer learning, multi-task learning, and multi-task learning with fine-tuning. CodeTrans outperforms the state-of-the-art models on all the tasks. To expedite future works in the software engineering domain, we have published our pre-trained models of CodeTrans. https://github.com/agemagician/CodeTrans, Comment: 28 pages, 6 tables and 1 figure

It is a long-lasting task to understand heat conduction phenomena beyond Fourier. Besides the low-temperature experiments on extremely pure crystals, it has turned out recently that heterogeneous materials with macro-scale size can also show thermal effects that cannot be modelled by the Fourier equation. This is called over-diffusive propagation, different from low-temperature observations, and is found in numerous samples made from metal foam, rocks, and composites. The measured temperature history is indeed similar to what Fourier's law predicts but the usual evaluation cannot provide reliable thermal parameters. This paper is a report on our experiments on several rock types, each type having multiple samples with different thicknesses. We show that size-dependent thermal behaviour can occur for both Fourier and non-Fourier situations. Moreover, based on the present experimental data, we find an empirical relation between the Fourier and non-Fourier parameters, which may be helpful in later experiments to develop a more robust and reliable evaluation procedure.

We construct a bi-Hamiltonian structure for the holomorphic spin Sutherland hierarchy based on collective spin variables. The construction relies on Poisson reduction of a bi-Hamiltonian structure on the holomorphic cotangent bundle of GL(n,C), which itself arises from the canonical symplectic structure and the Poisson structure of the Heisenberg double of the standard GL(n,C) Poisson--Lie group. The previously obtained bi-Hamiltonian structures of the hyperbolic and trigonometric real forms are recovered on real slices of the holomorphic spin Sutherland model., Comment: Expanded to 20 pages, contains a simplified formula of the second reduced Poisson bracket, more detailed derivations, and added references

The heat pulse (flash) experiment is a well-known and widely accepted method to measure the thermal diffusivity of a material. In recent years, it is observed that the thermal behavior of heterogeneous materials can show deviation from the classical Fourier equation, resulting in a different thermal diffusivity and requiring further thermal parameters to identify. Such heterogeneity can be inclusions in metal foams, layered structure in composites, or even cracks and porous parts in rocks. Furthermore, the next candidate, the so-called Guyer-Krumhansl equation, is tested on these experiments with success. However, these recent evaluations required a computationally intensive fitting procedure using countless numerical solutions, even when a good initial guess for the parameters is found by hand. This paper presents a Galerkin-type discretization for the Guyer-Krumhansl equation, which helped us find a reasonably simple analytical solution for time-dependent boundary conditions. Utilizing this analytical solution, we developed a new evaluation technique to immediately estimate all the necessary thermal parameters using the measured temperature history.

JOREK is a massively parallel fully implicit non-linear extended MHD code for realistic tokamak X-point plasmas. It has become a widely used versatile code for studying large-scale plasma instabilities and their control developed in an international community. This article gives a comprehensive overview of the physics models implemented, numerical methods applied for solving the equations and physics studies performed with the code. A dedicated section highlights some of the verification work done for the code. A hierarchy of different physics models is available including a free boundary and resistive wall extension and hybrid kinetic-fluid models. The code allows for flux-surface aligned iso-parametric finite element grids in single and double X-point plasmas which can be extended to the true physical walls and uses a robust fully implicit time stepping. Particular focus is laid on plasma edge and scrape-off layer (SOL) physics as well as disruption related phenomena. Among the key results obtained with JOREK regarding plasma edge and SOL, are deep insights into the dynamics of edge localized modes (ELMs), ELM cycles, and ELM control by resonant magnetic perturbations, pellet injection, as well as by vertical magnetic kicks. Also ELM free regimes, detachment physics, the generation and transport of impurities during an ELM, and electrostatic turbulence in the pedestal region are investigated. Regarding disruptions, the focus is on the dynamics of the thermal quench and current quench triggered by massive gas injection (MGI) and shattered pellet injection (SPI), runaway electron (RE) dynamics as well as the RE interaction with MHD modes, and vertical displacement events (VDEs). Also the seeding and suppression of tearing modes (TMs), the dynamics of naturally occurring thermal quenches triggered by locked modes, and radiative collapses are being studied., Comment: Comprehensive review published as special topic article in Nuclear Fusion

We present a numerical method for computing the evolution of a planar, star-shaped curve under a broad class of curvature-driven geometric flows, which we refer to as the Andrews-Bloore flows. This family of flows has two parameters that control one constant and one curvature-dependent component for the velocity in the direction of the normal to the curve. The Andrews-Bloore flow includes as special cases the well known Eikonal, curve-shortening and affine shortening flows, and for positive parameter values its evolution shrinks the area enclosed by the curve to zero in finite time. A question of key interest has been how the shape itself, as well as various shape descriptors evolve as this limit is approached. Star-like curves can be represented by a periodic scalar polar distance function $r(\varphi)$ measured from a reference point. An important question is how the numbers and the trajectories of critical points of the distance function $r(\varphi)$ and of the curvature $\kappa(\varphi)$ evolve under the Andrews-Bloore flows for different choices of the parameters. Our method is specifically designed to meet the challenge of computing accurate trajectories of these critical points up to the vicinity of a limiting shape. Each curve is represented by a piecewise polynomial periodic distance function, as determined by a chosen mesh; different types of meshes and mesh adaptation can be chosen to ensure a good balance between accuracy and computational cost. As we demonstrate with benchmark tests and two longer case studies, our method allows one to perform numerical investigations into subtle questions of planar curve evolution. More specifically we provide illustrations of some known results, numerical evidence for two stated conjectures, as well as new insights and observations regarding the limit shapes of initial curves with different types of symmetry.

We present the results of a single-pointing survey of 207 dense cores embedded in Planck Galactic Cold Clumps distributed in five different environments ($\lambda$ Orionis, Orion A, B, Galactic plane, and high latitudes) to identify dense cores on the verge of star formation for the study of the initial conditions of star formation. We observed these cores in eight molecular lines at 76-94 GHz using the Nobeyama 45-m telescope. We find that early-type molecules (e.g., CCS) have low detection rates and that late-type molecules (e.g., N$_2$H$^+$, c-C$_3$H$_2$) and deuterated molecules (e.g., N$_2$D$^+$, DNC) have high detection rates, suggesting that most of the cores are chemically evolved. The deuterium fraction (D/H) is found to decrease with increasing distance, indicating that it suffers from differential beam dilution between the D/H pair of lines for distant cores ($>$1 kpc). For $\lambda$ Orionis, Orion A, and B located at similar distances, D/H is not significantly different, suggesting that there is no systematic difference in the observed chemical properties among these three regions. We identify at least eight high D/H cores in the Orion region and two at high latitudes, which are most likely to be close to the onset of star formation. There is no clear evidence of the evolutionary change in turbulence during the starless phase, suggesting that the dissipation of turbulence is not a major mechanism for the beginning of star formation as judged from observations with a beam size of 0.04 pc., Comment: 55 pages, 12 Figures, 8 Tables, accepted for publication in ApJS

We report on the the trigonometric spin Ruijsenaars--Sutherland hierarchy derived recently by Poisson reduction of a bi-Hamiltonian hierarchy associated with free geodesic motion on the Lie group U(n). In particular, we give a direct proof of a previously stated result about the form of the second Poisson bracket in terms of convenient variables., Comment: Contribution to the proceedings of the XXXVIII Workshop on Geometric Methods in Physics (Bialowieza, June 2019), 9 pages

We investigate the trigonometric real form of the spin Ruijsenaars-Schneider system introduced, at the level of equations of motion, by Krichever and Zabrodin in 1995. This pioneering work and all earlier studies of the Hamiltonian interpretation of the system were performed in complex holomorphic settings; understanding the real forms is a non-trivial problem. We explain that the trigonometric real form emerges from Hamiltonian reduction of an obviously integrable 'free' system carried by a spin extension of the Heisenberg double of the ${\rm U}(n)$ Poisson-Lie group. The Poisson structure on the unreduced real phase space ${\rm GL}(n,\mathbb{C}) \times \mathbb{C}^{nd}$ is the direct product of that of the Heisenberg double and $d\geq 2$ copies of a ${\rm U}(n)$ covariant Poisson structure on $\mathbb{C}^n \simeq \mathbb{R}^{2n}$ found by Zakrzewski, also in 1995. We reduce by fixing a group valued moment map to a multiple of the identity, and analyze the resulting reduced system in detail. In particular, we derive on the reduced phase space the Hamiltonian structure of the trigonometric spin Ruijsenaars-Schneider system and we prove its degenerate integrability., Comment: 52 pages, 1 figure. v2: typos removed, final version

Computational biology and bioinformatics provide vast data gold-mines from protein sequences, ideal for Language Models taken from NLP. These LMs reach for new prediction frontiers at low inference costs. Here, we trained two auto-regressive models (Transformer-XL, XLNet) and four auto-encoder models (BERT, Albert, Electra, T5) on data from UniRef and BFD containing up to 393 billion amino acids. The LMs were trained on the Summit supercomputer using 5616 GPUs and TPU Pod up-to 1024 cores. Dimensionality reduction revealed that the raw protein LM-embeddings from unlabeled data captured some biophysical features of protein sequences. We validated the advantage of using the embeddings as exclusive input for several subsequent tasks. The first was a per-residue prediction of protein secondary structure (3-state accuracy Q3=81%-87%); the second were per-protein predictions of protein sub-cellular localization (ten-state accuracy: Q10=81%) and membrane vs. water-soluble (2-state accuracy Q2=91%). For the per-residue predictions the transfer of the most informative embeddings (ProtT5) for the first time outperformed the state-of-the-art without using evolutionary information thereby bypassing expensive database searches. Taken together, the results implied that protein LMs learned some of the grammar of the language of life. To facilitate future work, we released our models at https://github.com/agemagician/ProtTrans., Comment: 17 pages, 9 figures, 4 tables

We develop the theory of halving spaces to obtain lower bounds in real enumerative geometry. Halving spaces are topological spaces with an action of a Lie group $\Gamma$ with additional cohomological properties. For $\Gamma=\mathbb{Z}_2$ we recover the conjugation spaces of Hausmann, Holm and Puppe. For $\Gamma=\mathrm{U}(1)$ we obtain the circle spaces. We show that real even and quaternionic partial flag manifolds are circle spaces leading to non-trivial lower bounds for even real and quaternionic Schubert problems. To prove that a given space is a halving space, we generalize results of Borel and Haefliger on the cohomology classes of real subvarieties and their complexifications. The novelty is that we are able to obtain results in rational cohomology instead of modulo 2. The equivariant extension of the theory of circle spaces leads to generalizations of the results of Borel and Haefliger on Thom polynomials., Comment: 30 pages

We study motivic Chern classes of cones. First we show examples of projective cones of smooth curves such that their various $K$-classes (sheaf theoretic, push-forward and motivic) are all different. Then we show connections between the torus equivariant motivic Chern class of a projective variety and of its affine cone, generalizing results on projective Thom polynomials., Comment: 18 pages

FU Orionis objects (FUors) are rapidly-accreting, pre-main sequence objects that are known to exhibit large outbursts at optical and near-infrared wavelengths, with post-eruption, small-scale photometric variability superimposed on longer-term trends. In contrast, little is known about the variability of FUors at longer wavelengths. To explore this further, we observed six FUor objects using the NOrthern Extended Millimeter Array (NOEMA) and for a subset of three objects we obtained coordinated observations with NOEMA and the Lowell Discovery Telescope (LDT). In combination with previously published NOEMA observations from 2014, our 2017 observations of V1735 Cyg provide the first detection of variability in an FUor object at 2.7 mm. In the absence of significant optical variability, we discount the possibility that the mm flux density changed as a result of irradiation from the central disk. In addition, a change in the dust mass due to infall is highly unlikely. A plausible explanation for the change in 2.7 mm flux density is variability in free-free emission due to changes in the object's jet/wind. Thus, it may be that free-free emission in some FUor objects is significant at $\sim$3 mm and must be considered when deriving disk masses in order to help constrain the mechanism responsible for triggering FUor outbursts., Comment: 13 pages, 8 figures, to be published in The Astrophysical Journal

With the increased expectation of artificial intelligence, academic research face complex questions of human-centred, responsible and trustworthy technology embedded into society and culture. Several academic debates, social consultations and impact studies are available to reveal the key aspects of the changing human-machine ecosystem. To contribute to these studies, hundreds of related academic sources are summarized below regarding AI-driven decisions and valuable AI. In details, sociocultural filters, taxonomy of human-machine decisions and perspectives of value-based AI are in the focus of this literature review. For better understanding, it is proposed to invite stakeholders in the prepared large-scale survey about the next generation AI that investigates issues that go beyond the technology., Comment: 5 pages, 2 figures

The purpose is to investigate the expected participation and mentality of smart citizens in smart cities. The key question is the role of the human factor in smart environments globally studied through a research corpus of 150 documents including mainstream summaries trend reports, white papers and visions of business, governmental and university research cooperations reaching a wide audience of the subject. Foremost, a short review of the changing scholarly trends is presented as a theoretical framework. Concerning its key ideas, the corpus based findings are recapped and analysed by content networks and the most referred city strategies. Besides, a critical approach reveal further required factors and risks to investigate. The ultimate goal is to understand how the smart city landscape is shaped by citizen-based strategies, open data, empowerment and responsibility. Accordingly, the paper closes with further considerations regarding the importance of anonymous open data, advantages of neighbourhood-based implementations, aspects of permanent and temporary citizen-engagements, interpretation of metaphors or upcoming technologies, and also, privacy and ethical issues. The results provide the policy development and the emerging scholarly interest with a framework study., Comment: 18 pages, 4 figures