Carnegie Mellon University

About: Carnegie Mellon University is a education organization based out in Pittsburgh, Pennsylvania, United States. It is known for research contribution in the topics: Population & Robot. The organization has 36317 authors who have published 104359 publications receiving 5975734 citations. The organization is also known as: CMU & Carnegie Mellon.

Composite Quasar Spectra from the Sloan Digital Sky Survey

Abstract: We have created a variety of composite quasar spectra using a homogeneous data set of over 2200 spectra from the Sloan Digital Sky Survey (SDSS). The quasar sample spans a redshift range of 0.044 ≤ z ≤ 4.789 and an absolute r' magnitude range of -18.0 to -26.5. The input spectra cover an observed wavelength range of 3800–9200 A at a resolution of 1800. The median composite covers a rest-wavelength range from 800 to 8555 A and reaches a peak signal-to-noise ratio of over 300 per 1 A resolution element in the rest frame. We have identified over 80 emission-line features in the spectrum. Emission-line shifts relative to nominal laboratory wavelengths are seen for many of the ionic species. Peak shifts of the broad permitted and semiforbidden lines are strongly correlated with ionization energy, as previously suggested, but we find that the narrow forbidden lines are also shifted by amounts that are strongly correlated with ionization energy. The magnitude of the forbidden line shifts is 100 km s-1, compared with shifts of up to 550 km s-1 for some of the permitted and semiforbidden lines. At wavelengths longer than the Lyα emission, the continuum of the geometric mean composite is well fitted by two power laws, with a break at ≈5000 A. The frequency power-law index, αν, is -0.44 from ≈1300 to 5000 A and -2.45 redward of ≈5000 A. The abrupt change in slope can be accounted for partly by host-galaxy contamination at low redshift. Stellar absorption lines, including higher order Balmer lines, seen in the composites suggest that young or intermediate-age stars make a significant contribution to the light of the host galaxies. Most of the spectrum is populated by blended emission lines, especially in the range 1500–3500 A, which can make the estimation of quasar continua highly uncertain unless large ranges in wavelength are observed. An electronic table of the median quasar template is available.

A Unified, Merger-driven Model of the Origin of Starbursts, Quasars, the Cosmic X-Ray Background, Supermassive Black Holes, and Galaxy Spheroids

Abstract: We present an evolutionary model for starbursts, quasars, and spheroidal galaxies in which mergers between gas-rich galaxies drive nuclear inflows of gas, producing starbursts and feeding the buried growth of supermassive black holes (BHs) until feedback expels gas and renders a briefly visible optical quasar. The quasar lifetime and obscuring column density depend on both the instantaneous and peak quasar luminosity, and we determine this dependence using a large set of galaxy merger simulations varying galaxy properties, orbital geometry, and gas physics. We use these fits to deconvolve observed quasar luminosity functions and obtain the evolution of the formation rate of quasars with peak luminosity, (Lpeak, z). Quasars spend extended periods at luminosities well below peak, so (Lpeak) has a maximum corresponding to the break in the observed luminosity function. From (Lpeak) and our simulations, we obtain self-consistent hard and soft X-ray and optical luminosity functions and predict many observables at multiple redshifts, including column density distributions of optical and X-ray samples, the luminosity function of broad-line quasars in X-ray samples and broad-line fraction versus luminosity, active BH mass functions, the distribution of Eddington ratios, the mass function of relic BHs and total BH mass density, and the cosmic X-ray background. In every case, our predictions agree well with observed estimates, without invoking ad hoc assumptions about source properties or distributions. We provide a library of Monte Carlo realizations of our models for comparison with observations.

Instabilities and pattern formation in crystal growth

Abstract: Several common modes of crystal growth provide particularly simple and elegant examples of spontaneous pattern formation in nature. Phenomena of interest here are those in which an advancing nonfaceted solidification front suffers an instability and subsequently reorganizes itself into a more complex mode of behavior. The purpose of this essay is to examine several such situations and, in doing this, to identify a few new theoretical ideas and a larger number of outstanding problems. The systems studied are those in which solidification is controlled entirely by a single diffusion process, either the flow of latent heat away from a moving interface or the analogous redistribution of chemical constituents. Convective effects are ignored, as are most effects of crystalline anisotropy. The linear theory of the Mullins-Sekerka instability is reviewed for simple planar and spherical cases and also for a special model of directional solidification. These techniques are then extended to the case of a freely growing dendrite, and it is shown how this analysis leads to an understanding of sidebranching and tip-splitting instabilities. A marginal-stability hypothesis is introduced; and it is argued that this intrinsically nonlinear theory, if valid, permits aone to use results of linear-stability analysis to predict dendritic growth rates. The review concludes with a discussion of nonlinear effects in directional solidication. The nonplanar, cellular interfaces which emerge in this situation have much in common with convection patterns in hydrodynamics. The cellular stability problem is discussed briefly, and some preliminary attempts to do calculations in the strongly nonlinear regime are summarized.

Accuracy of AH n equilibrium geometries by single determinant molecular orbital theory

Abstract: A simple level of ab initio molecular orbital theory with a split-valence shell basis with d-type polarization functions (6–31G*) is used to predict equilibrium geometries for the ground and some low-lying excited states of AHn molecules and cations where A is carbon, nitrogen, oxygen or fluorine. The results are shown to be close to the limit for single determinant wave functions in cases where corresponding computations with more extensive bases are available. Comparison with experimental results also shows good agreement although a systematic underestimation of bond lengths up to 3 per cent is evident. For systems where no experimental data are available, the results provide predictions of equilibrium geometry.

Revisiting Unreasonable Effectiveness of Data in Deep Learning Era

Abstract: The success of deep learning in vision can be attributed to: (a) models with high capacity; (b) increased computational power; and (c) availability of large-scale labeled data. Since 2012, there have been significant advances in representation capabilities of the models and computational capabilities of GPUs. But the size of the biggest dataset has surprisingly remained constant. What will happen if we increase the dataset size by 10 × or 100 × ? This paper takes a step towards clearing the clouds of mystery surrounding the relationship between ‘enormous data’ and visual deep learning. By exploiting the JFT-300M dataset which has more than 375M noisy labels for 300M images, we investigate how the performance of current vision tasks would change if this data was used for representation learning. Our paper delivers some surprising (and some expected) findings. First, we find that the performance on vision tasks increases logarithmically based on volume of training data size. Second, we show that representation learning (or pre-training) still holds a lot of promise. One can improve performance on many vision tasks by just training a better base model. Finally, as expected, we present new state-of-the-art results for different vision tasks including image classification, object detection, semantic segmentation and human pose estimation. Our sincere hope is that this inspires vision community to not undervalue the data and develop collective efforts in building larger datasets.