-
DINO as a von Mises-Fisher mixture model
Authors:
Hariprasath Govindarajan,
Per Sidén,
Jacob Roll,
Fredrik Lindsten
Abstract:
Self-distillation methods using Siamese networks are popular for self-supervised pre-training. DINO is one such method based on a cross-entropy loss between $K$-dimensional probability vectors, obtained by applying a softmax function to the dot product between representations and learnt prototypes. Given the fact that the learned representations are $L^2$-normalized, we show that DINO and its deri…
▽ More
Self-distillation methods using Siamese networks are popular for self-supervised pre-training. DINO is one such method based on a cross-entropy loss between $K$-dimensional probability vectors, obtained by applying a softmax function to the dot product between representations and learnt prototypes. Given the fact that the learned representations are $L^2$-normalized, we show that DINO and its derivatives, such as iBOT, can be interpreted as a mixture model of von Mises-Fisher components. With this interpretation, DINO assumes equal precision for all components when the prototypes are also $L^2$-normalized. Using this insight we propose DINO-vMF, that adds appropriate normalization constants when computing the cluster assignment probabilities. Unlike DINO, DINO-vMF is stable also for the larger ViT-Base model with unnormalized prototypes. We show that the added flexibility of the mixture model is beneficial in terms of better image representations. The DINO-vMF pre-trained model consistently performs better than DINO on a range of downstream tasks. We obtain similar improvements for iBOT-vMF vs iBOT and thereby show the relevance of our proposed modification also for other methods derived from DINO.
△ Less
Submitted 17 May, 2024;
originally announced May 2024.
-
Creating a three dimensional intrinsic electric dipole on rotated CrI$_3$ bilayers
Authors:
Shiva P. Poudel,
Juan M. Marmolejo-Tejada,
Joseph E. Roll,
Martín A. Mosquera,
Salvador Barraza-Lopez
Abstract:
Two-dimensional (2D) materials are being explored as a novel multiferroic platform. One of the most studied magnetoelectric multiferroic 2D materials are antiferromagnetically-coupled (AFM) CrI$_3$ bilayers. Neglecting magnetism, those bilayers possess a crystalline point of inversion, which is only removed by the antiparallel spin configuration among its two constituent monolayers. The resultant…
▽ More
Two-dimensional (2D) materials are being explored as a novel multiferroic platform. One of the most studied magnetoelectric multiferroic 2D materials are antiferromagnetically-coupled (AFM) CrI$_3$ bilayers. Neglecting magnetism, those bilayers possess a crystalline point of inversion, which is only removed by the antiparallel spin configuration among its two constituent monolayers. The resultant intrinsic electric dipole on those bilayers has a magnitude no larger than 0.04 pC/m, it points out-of-plane, and it reverts direction when the--Ising-like--cromium spins are flipped (toward opposite layers {\em versus} away from opposite layers). The combined presence of antiferromagnetism and a weak intrinsic electric dipole makes this material a two-dimensional magnetoelectric multiferroic. Here, we remove the crystalline center of inversion of the bilayer by a relative $60^{\circ}$ rotation of its constituent monolayers. This process {\em enhances} the out-of-plane intrinsic electric dipole tenfold with respect to its magnitude in the non-rotated AFM bilayer and also creates an even stronger and switchable in-plane intrinsic electric dipole. The ability to create a three-dimensional electric dipole is important, because it enhances the magnetoelectric coupling on this experimentally accessible 2D material, which is explicitly calculated here as well.
△ Less
Submitted 1 May, 2023;
originally announced May 2023.
-
Slippery paraelectric transition metal dichalcogenide bilayers
Authors:
Juan M. Marmolejo-Tejada,
Joseph E. Roll,
Shiva Prasad Poudel,
Salvador Barraza-Lopez,
Martin A. Mosquera
Abstract:
Traditional ferroelectrics undergo thermally-induced phase transitions whereby their structural symmetry increases. The associated higher-symmetry structure is dubbed {\em paraelectric}. Ferroelectric transition metal dichalcogenide bilayers have been recently shown to become paraelectric, but not much has been said of the atomistic configuration of such a phase. As discovered through numerical ca…
▽ More
Traditional ferroelectrics undergo thermally-induced phase transitions whereby their structural symmetry increases. The associated higher-symmetry structure is dubbed {\em paraelectric}. Ferroelectric transition metal dichalcogenide bilayers have been recently shown to become paraelectric, but not much has been said of the atomistic configuration of such a phase. As discovered through numerical calculations that include molecular dynamics here, their paraelectricity can only be ascribed to a time average of ferroelectric phases with opposing intrinsic polarizations, whose switching requires macroscopically large areas to slip in unison.
△ Less
Submitted 28 September, 2022;
originally announced September 2022.
-
Elasticity of 2D ferroelectrics across their paraelectric phase transformation
Authors:
Joseph E. Roll,
John M. Davis,
John W. Villanova,
Salvador Barraza-Lopez
Abstract:
The mechanical behavior of two-dimensional (2D) materials across 2D phase changes is unknown, and the finite temperature ($T$) elasticity of paradigmatic SnSe monolayers -- ferroelectric 2D materials turning paraelectric as their unit cell (u.c.) turns from a rectangle onto a square -- is described here in a progressive manner. To begin with, their zero$-T$ {\em elastic energy landscape} gives way…
▽ More
The mechanical behavior of two-dimensional (2D) materials across 2D phase changes is unknown, and the finite temperature ($T$) elasticity of paradigmatic SnSe monolayers -- ferroelectric 2D materials turning paraelectric as their unit cell (u.c.) turns from a rectangle onto a square -- is described here in a progressive manner. To begin with, their zero$-T$ {\em elastic energy landscape} gives way to (Boltzmann-like) averages from which the elastic behavior is determined. These estimates are complemented with results from the strain-fluctuation method, which employs the energy landscape or {\em ab initio} molecular dynamics (MD) data. Both approaches capture the coalescence of elastic moduli $\langle C_{11}(T)\rangle=\langle C_{22}(T)\rangle$ due to the structural transformation. The broad evolution and sudden changes of elastic parameters $\langle C_{11}(T)\rangle$, $\langle C_{22}(T)\rangle$, and $\langle C_{12}(T)\rangle$ of these atomically-thin phase-change membranes establishes a heretofore overlooked connection among 2D materials and soft matter.
△ Less
Submitted 15 March, 2022;
originally announced March 2022.
-
Magnetic Topological Semimetal Phase with Electronic Correlation Enhancement in SmSbTe
Authors:
Krishna Pandey,
Debashis Mondal,
John William Villanova,
Joseph Roll,
Rabindra Basnet,
Aaron Wegner,
Gokul Acharya,
Md Rafique Un Nabi,
Barun Ghosh,
Jun Fujii,
Jian Wang,
Bo Da,
Amit Agarwal,
Ivana Vobornik,
Antonio Politano,
Salvador Barraza-Lopez,
Jin Hu
Abstract:
The ZrSiS family of compounds hosts various exotic quantum phenomena due to the presence of both topological nonsymmorphic Dirac fermions and nodal-line fermions. In this material family, the LnSbTe (Ln= lanthanide) compounds are particularly interesting owing to the intrinsic magnetism from magnetic Ln which leads to new properties and quantum states. In this work, the authors focus on the previo…
▽ More
The ZrSiS family of compounds hosts various exotic quantum phenomena due to the presence of both topological nonsymmorphic Dirac fermions and nodal-line fermions. In this material family, the LnSbTe (Ln= lanthanide) compounds are particularly interesting owing to the intrinsic magnetism from magnetic Ln which leads to new properties and quantum states. In this work, the authors focus on the previously unexplored compound SmSbTe. The studies reveal a rare combination of a few functional properties in this material, including antiferromagnetism with possible magnetic frustration, electron correlation enhancement, and Dirac nodal-line fermions. These properties enable SmSbTe as a unique platform to explore exotic quantum phenomena and advanced functionalities arising from the interplay between magnetism, topology, and electronic correlations.
△ Less
Submitted 24 August, 2021;
originally announced August 2021.
-
Binospec Software System
Authors:
Jan Kansky,
Igor Chilingarian,
Daniel Fabricant,
Anne Matthews,
Sean Moran,
Martin Paegert,
J. Duane Gibson,
Dallan Porter,
John Roll
Abstract:
Binospec is a high-throughput, 370 to 1000 nm, imaging spectrograph that addresses two adjacent 8' by 15' fields of view. Binospec was commissioned in late 2017 at the f/5 focus of the 6.5m MMT and is now available to all MMT observers. Here we describe the Binospec software used for observation planning, instrument control, and data reduction. The software and control systems incorporate a high l…
▽ More
Binospec is a high-throughput, 370 to 1000 nm, imaging spectrograph that addresses two adjacent 8' by 15' fields of view. Binospec was commissioned in late 2017 at the f/5 focus of the 6.5m MMT and is now available to all MMT observers. Here we describe the Binospec software used for observation planning, instrument control, and data reduction. The software and control systems incorporate a high level of automation to minimize observer workload. Instrument configuration and observation sequencing is implemented using a database-driven approach to maximize observatory efficiency. A web-based interface allows users to define observations, monitor status, and retrieve data products.
△ Less
Submitted 8 May, 2019;
originally announced May 2019.
-
Evaluating model calibration in classification
Authors:
Juozas Vaicenavicius,
David Widmann,
Carl Andersson,
Fredrik Lindsten,
Jacob Roll,
Thomas B. Schön
Abstract:
Probabilistic classifiers output a probability distribution on target classes rather than just a class prediction. Besides providing a clear separation of prediction and decision making, the main advantage of probabilistic models is their ability to represent uncertainty about predictions. In safety-critical applications, it is pivotal for a model to possess an adequate sense of uncertainty, which…
▽ More
Probabilistic classifiers output a probability distribution on target classes rather than just a class prediction. Besides providing a clear separation of prediction and decision making, the main advantage of probabilistic models is their ability to represent uncertainty about predictions. In safety-critical applications, it is pivotal for a model to possess an adequate sense of uncertainty, which for probabilistic classifiers translates into outputting probability distributions that are consistent with the empirical frequencies observed from realized outcomes. A classifier with such a property is called calibrated. In this work, we develop a general theoretical calibration evaluation framework grounded in probability theory, and point out subtleties present in model calibration evaluation that lead to refined interpretations of existing evaluation techniques. Lastly, we propose new ways to quantify and visualize miscalibration in probabilistic classification, including novel multidimensional reliability diagrams.
△ Less
Submitted 19 February, 2019;
originally announced February 2019.
-
Leveraging Historical Associations between Requirements and Source Code to Identify Impacted Classes
Authors:
Davide Falessi,
Justin Roll,
Jin Guo,
Jane Cleland-Huang
Abstract:
As new requirements are introduced and implemented in a software system, developers must identify the set of source code classes which need to be changed. Therefore, past effort has focused on predicting the set of classes impacted by a requirement. In this paper, we introduce and evaluate a new type of information based on the intuition that the set of requirements which are associated with histo…
▽ More
As new requirements are introduced and implemented in a software system, developers must identify the set of source code classes which need to be changed. Therefore, past effort has focused on predicting the set of classes impacted by a requirement. In this paper, we introduce and evaluate a new type of information based on the intuition that the set of requirements which are associated with historical changes to a specific class are likely to exhibit semantic similarity to new requirements which impact that class. This new Requirements to Requirements Set (R2RS) family of metrics captures the semantic similarity between a new requirement and the set of existing requirements previously associated with a class. The aim of this paper is to present and evaluate the usefulness of R2RS metrics in predicting the set of classes impacted by a requirement. We consider 18 different R2RS metrics by combining six natural language processing techniques to measure the semantic similarity among texts (e.g., VSM) and three distribution scores to compute overall similarity (e.g., average among similarity scores). We evaluate if R2RS is useful for predicting impacted classes in combination and against four other families of metrics that are based upon temporal locality of changes, direct similarity to code, complexity metrics, and code smells. Our evaluation features five classifiers and 78 releases belonging to four large open-source projects, which result in over 700,000 candidate impacted classes. Experimental results show that leveraging R2RS information increases the accuracy of predicting impacted classes practically by an average of more than 60% across the various classifiers and projects.
△ Less
Submitted 20 August, 2018;
originally announced August 2018.
-
Detection of Another Molecular Bubble in the Galactic Center
Authors:
Shiho Tsujimoto,
Tomoharu Oka,
Shunya Takekawa,
Masaya Yamada,
Sekito Tokuyama,
Yuhei Iwata,
Justin A. Roll
Abstract:
The $l=-1.2^{\circ}$ region in the Galactic center has a high CO $J$=3-2/$J$=1-0 intensity ratio and extremely broad velocity width. This paper reports the detection of five expanding shells in the $l=-1.2^{\circ}$ region based on the CO $J$=1-0, $^{13}$CO $J$=1-0, CO $J$=3-2, and SiO $J$=8-7 line data sets obtained with the Nobeyama Radio Observatory 45 m telescope and James Clerk Maxwell Telesco…
▽ More
The $l=-1.2^{\circ}$ region in the Galactic center has a high CO $J$=3-2/$J$=1-0 intensity ratio and extremely broad velocity width. This paper reports the detection of five expanding shells in the $l=-1.2^{\circ}$ region based on the CO $J$=1-0, $^{13}$CO $J$=1-0, CO $J$=3-2, and SiO $J$=8-7 line data sets obtained with the Nobeyama Radio Observatory 45 m telescope and James Clerk Maxwell Telescope. The kinetic energy and expansion time of the expanding shells are estimated to be $10^{48.3-50.8}$ erg and $10^{4.7-5.0}$ yr, respectively. The origin of these expanding shells is discussed. The total kinetic energy of $10^{51}$ erg and the typical expansion time of ${\sim}10^5$ yr correspond to multiple supernova explosions at a rate of $10^{-5}$-$10^{-4}$ yr$^{-1}$. This indicates that the $l=-1.2^{\circ}$ region may be a molecular bubble associated with an embedded massive star cluster, although the absence of an infrared counterpart makes this interpretation somewhat controversial. The expansion time of the shells increases as the Galactic longitude decreases, suggesting that the massive star cluster is moving from Galactic west to east with respect to the interacting molecular gas. We propose a model wherein the cluster is moving along the innermost x1 orbit and the interacting gas collides with it from the Galactic eastern side.
△ Less
Submitted 19 March, 2018;
originally announced March 2018.
-
Data reduction pipeline for the MMT Magellan Infrared Spectrograph
Authors:
Igor Chilingarian,
Warren Brown,
Daniel Fabricant,
Brian McLeod,
John Roll,
Andrew Szentgyorgyi
Abstract:
We describe principal components of the new spectroscopic data pipeline for the multi-object MMT/Magellan Infrared Spectrograph (MMIRS). The pipeline is implemented in IDL and C++. The performance of the data processing algorithms is sufficient to reduce a single dataset in 2--3 min on a modern PC workstation so that one can use the pipeline as a quick-look tool during observations. We provide an…
▽ More
We describe principal components of the new spectroscopic data pipeline for the multi-object MMT/Magellan Infrared Spectrograph (MMIRS). The pipeline is implemented in IDL and C++. The performance of the data processing algorithms is sufficient to reduce a single dataset in 2--3 min on a modern PC workstation so that one can use the pipeline as a quick-look tool during observations. We provide an example of the spectral data processed by our pipeline and demonstrate that the sky subtraction quality gets close to the limits set by the Poisson photon statistics.
△ Less
Submitted 16 December, 2012;
originally announced December 2012.
-
MMT & Magellan Infrared Spectrograph
Authors:
Brian McLeod,
Daniel Fabricant,
George Nystrom,
Ken McCracken,
Stephen Amato,
Henry Bergner,
Warren Brown,
Michael Burke,
Igor Chilingarian,
Maureen Conroy,
Dylan Curley,
Gabor Furesz,
John Geary,
Edward Hertz,
Justin Holwell,
Anne Matthews,
Tim Norton,
Sang Park,
John Roll,
Joseph Zajac,
Harland Epps,
Paul Martini
Abstract:
The MMT and Magellan infrared spectrograph (MMIRS) is a cryogenic multiple slit spectrograph operating in the wavelength range 0.9-2.4 micron. MMIRS' refractive optics offer a 6.9 by 6.9 arcmin field of view for imaging with a spatial resolution of 0.2 arcsec per pixel on a HAWAII-2 array. For spectroscopy, MMIRS can be used with long slits up to 6.9 arcmin long, or with custom slit masks having s…
▽ More
The MMT and Magellan infrared spectrograph (MMIRS) is a cryogenic multiple slit spectrograph operating in the wavelength range 0.9-2.4 micron. MMIRS' refractive optics offer a 6.9 by 6.9 arcmin field of view for imaging with a spatial resolution of 0.2 arcsec per pixel on a HAWAII-2 array. For spectroscopy, MMIRS can be used with long slits up to 6.9 arcmin long, or with custom slit masks having slitlets distributed over a 4 by 6.9 arcmin area. A range of dispersers offer spectral resolutions of 800 to 3000. MMIRS is designed to be used at the f/5 foci of the MMT or Magellan Clay 6.5m telescopes. MMIRS was commissioned in 2009 at the MMT and has been in routine operation at the Magellan Clay Telescope since 2010. MMIRS is being used for a wide range of scientific investigations from exoplanet atmospheres to Ly-alpha emitters.
△ Less
Submitted 26 November, 2012;
originally announced November 2012.
-
Kinematics of NGC 2264: signs of cluster formation
Authors:
Gabor Furesz,
Lee W. Hartmann,
Andrew H. Szentgyorgyi,
Naomi A. Ridge,
Luisa Rebull,
John Stauffer,
David W. Latham,
Maureen A. Conroy,
Daniel G. Fabricant,
John Roll
Abstract:
We present results from 1078 high resolution spectra of 990 stars in the young open cluster NGC 2264, obtained with the Hectochelle multiobject echelle spectrograph on the 6.5m MMT. We confirm 471 stars as members, based on their radial velocity and/or H-alpha emission. The radial velocity distribution of cluster members is non-Gaussian with a dispersion of approx 3.5 km/s. We find a substantial…
▽ More
We present results from 1078 high resolution spectra of 990 stars in the young open cluster NGC 2264, obtained with the Hectochelle multiobject echelle spectrograph on the 6.5m MMT. We confirm 471 stars as members, based on their radial velocity and/or H-alpha emission. The radial velocity distribution of cluster members is non-Gaussian with a dispersion of approx 3.5 km/s. We find a substantial north-south velocity gradient and spatially coherent structure inthe radial velocity distribution, similar to that seen in the molecular gas in the region. Our results suggest that there are at least three distinguishable subclusters in NGC 2264, correlated with similar structure seen in 13CO emission, which is likely to be a remnant of initial structure in this very young cluster. We propose that this substructure is the result of gravitational amplification of initial inhomogeneities during overall collapse to a filamentary distribution of gas and stars, as found in simulations by Burkert & Hartman (2004).
△ Less
Submitted 5 November, 2007; v1 submitted 2 November, 2007;
originally announced November 2007.
-
Hectospec, the MMT's 300 Optical Fiber-Fed Spectrograph
Authors:
Daniel Fabricant,
Robert Fata,
John Roll,
Edward Hertz,
Nelson Caldwell,
Thomas Gauron,
John Geary,
Brian McLeod,
Andrew Szentgyorgyi
Abstract:
The Hectospec is a 300 optical fiber fed spectrograph commissioned at the MMT in the spring of 2004. A pair of high-speed six-axis robots move the 300 fiber buttons between observing configurations within ~300 s and to an accuracy ~25 microns. The optical fibers run for 26 m between the MMT's focal surface and the bench spectrograph operating at R~1000-2000. Another high dispersion bench spectro…
▽ More
The Hectospec is a 300 optical fiber fed spectrograph commissioned at the MMT in the spring of 2004. A pair of high-speed six-axis robots move the 300 fiber buttons between observing configurations within ~300 s and to an accuracy ~25 microns. The optical fibers run for 26 m between the MMT's focal surface and the bench spectrograph operating at R~1000-2000. Another high dispersion bench spectrograph offering R~5,000, Hectochelle, is also available. The system throughput, including all losses in the telescope optics, fibers, and spectrograph peaks at ~10% at the grating blaze in 1" FWHM seeing. Correcting for aperture losses at the 1.5" diameter fiber entrance aperture, the system throughput peaks at $\sim$17%. Hectospec has proven to be a workhorse instrument at the MMT. Hectospec and Hectochelle together were scheduled for 1/3 of the available nights since its commissioning. Hectospec has returned \~60,000 reduced spectra for 16 scientific programs during its first year of operation.
△ Less
Submitted 25 August, 2005;
originally announced August 2005.
-
Surveying the Inner Halo of the Galaxy with 2MASS-Selected Horizontal Branch Candidates
Authors:
Warren R. Brown,
M. J. Geller,
S. J. Kenyon,
T. C. Beers,
M. J. Kurtz,
J. B. Roll
Abstract:
We use 2MASS photometry to select blue horizontal branch (BHB) candidates covering the sky |b|>15 deg. A 12.5<J<15.5 sample of BHB stars traces the thick disk and inner halo to d<9 kpc, with a density comparable to that of M giant stars. We base our sample selection strategy on the Century Survey Galactic Halo Project, a survey that provides a complete, spectroscopically-identified sample of blu…
▽ More
We use 2MASS photometry to select blue horizontal branch (BHB) candidates covering the sky |b|>15 deg. A 12.5<J<15.5 sample of BHB stars traces the thick disk and inner halo to d<9 kpc, with a density comparable to that of M giant stars. We base our sample selection strategy on the Century Survey Galactic Halo Project, a survey that provides a complete, spectroscopically-identified sample of blue stars to a similar depth as the 2MASS catalog. We show that a -0.20<(J-H)_0<0.10, -0.10<(H-K)_0<0.10 color-selected sample of stars is 65% complete for BHB stars, and is composed of 47% BHB stars. We apply this photometric selection to the full 2MASS catalog, and see no spatial overdensities of BHB candidates at high Galactic latitude |b|>50 deg. We insert simulated star streams into the data and conclude that the high Galactic latitude BHB candidates are consistent with having no ~5 deg wide star stream with density greater than 0.33 objects deg^-2 at the 95% confidence level. The absence of structure suggests there have been no major accretion events in the inner halo in the last few Gyr. However, at low Galactic latitudes a two-point angular correlation analysis reveals structure on angular scales <1 deg. This structure is apparently associated with stars in the thick disk, and has a physical scale of 10-100 pc. Interestingly, such structures are expected by cosmological simulations that predict the majority of the thick disk may arise from accretion and disruption of satellite mergers.
△ Less
Submitted 1 December, 2003; v1 submitted 29 September, 2003;
originally announced September 2003.
-
The Redshift of the Optical Transient Associated with GRB 010222
Authors:
Saurabh Jha,
Michael A. Pahre,
Peter M. Garnavich,
Michael L. Calkins,
Roy E. Kilgard,
Thomas Matheson,
Jonathan C. McDowell,
John B. Roll,
Krzysztof Z. Stanek
Abstract:
The gamma-ray burst (GRB) 010222 is the brightest GRB detected to date by the BeppoSAX satellite. Prompt identification of the associated optical transient (OT) allowed for spectroscopy with the Tillinghast 1.5m telescope at F. L. Whipple Observatory while the source was still relatively bright (R ~ 18.6 mag), within five hours of the burst. The OT shows a blue continuum with many superimposed a…
▽ More
The gamma-ray burst (GRB) 010222 is the brightest GRB detected to date by the BeppoSAX satellite. Prompt identification of the associated optical transient (OT) allowed for spectroscopy with the Tillinghast 1.5m telescope at F. L. Whipple Observatory while the source was still relatively bright (R ~ 18.6 mag), within five hours of the burst. The OT shows a blue continuum with many superimposed absorption features corresponding to metal lines at z = 1.477, 1.157, and possibly also at 0.928. The redshift of GRB 010222 is therefore unambiguously placed at z >= 1.477. The high number of Mg II absorbers and especially the large equivalent widths of the Mg II, Mg I, and Fe II absorption lines in the z = 1.477 system further argue either for a very small impact parameter or that the z = 1.477 system is the GRB host galaxy itself. The spectral index of the OT is relatively steep, beta = 0.89 +/- 0.03, and this cannot be caused by dust with a standard Galactic extinction law in the z = 1.477 absorption system. This spectroscopic identification of the redshift of GRB 010222 shows that prompt and well-coordinated followup of bright GRBs can be successful even with telescopes of modest aperture.
△ Less
Submitted 16 May, 2001; v1 submitted 5 March, 2001;
originally announced March 2001.