Automatic Determination of Stellar Atmospheric Parameters Using Neural Networks and Instance-Based Machine Learning

In this article we show how machine learning methods can beeffectively applied to the problem of automatically predictingstellar atmospheric parameters from spectral information, a veryimportant problem in stellar astronomy. We apply feedforwardneural networks, Kohonen's self-organizing maps andlocally-weighted regression to predict the stellar atmosphericparameters effective temperature, surface gravity and metallicityfrom spectral indices. Our experimental results show that thethree methods are capable of predicting the parameters with verygood accuracy. Locally weighted regression gives slightly betterresults than the other methods using the original dataset asinput, while self-organizing maps outperform the other methods when significant amounts of noise are added. We also implemented a heterogeneous ensemble of predictors, combining the results given by the three algorithms. This ensemble yields better results than any of the three algorithms alone, using both the original and the noisy data.     

Parameter information from non-linear cosmological fields

We develop a general formalism for analysing parameter information from non-Gaussian cosmic fields. The method can be adapted to include the non-linear effects in galaxy redshift surveys, weak lensing surveys and cosmic velocity field surveys as part of parameter estimation. It can also be used as a test of non-Gaussianity of the cosmic microwave background. Generalizing maximum-likelihood analysis to second order, we calculate the non-linear Fisher information matrix and likelihood surfaces in parameter space. To this order we find that the information content is always increased by including non-linearity. Our methods are applied to a realistic model of a galaxy redshift survey, including non-linear evolution, galaxy bias, shot-noise and redshift-space distortions to second order. We find that including non-linearities allows all of the degeneracies between parameters to be lifted. Marginalized parameter uncertainties of a few per cent will then be obtainable using forthcoming galaxy redshift surveys.     

Multiple-source localization using GPS technology and receivedarrival time structure analysis in an air-deployed system

An efficient and robust method has been developed to locate multiple impulsive sources in an ocean environment. Global position system (GPS) receivers were installed on sonobuoys to obtain their locations within a few meters of accuracy. A sonobuoy field was deployed in a ring-type pattern. Charges were then set off at arbitrary locations within the ring, High-resolution plots were used to obtain direct path and/or first bottom bounce arrivals on each buoy. A model grid of arrival times was constructed, corresponding to the dimensions of the buoy field. A ray model previously developed here at the Applied Research Laboratories at the University of Texas at Austin (ARL:UT) was used to obtain model travel times. The minimum value of the least-square-type error between the real arrival times and the modeled travel times resulted in an unambiguous location of the source, within the limits of the grid spacing chosen. This value was calculated by picking one receiver as the reference and then summing the timing errors of the remaining receivers relative to the reference. Successive iterations with finer grid spacings result in source localization within the accuracy of the buoy locations. The localization routine was extended by allowing permutations of the pulse arrivals on each buoy to account for multiple sources closely separated in time and/or space. An automated correlation technique is presented as an alternative to the leading edge-detection method used here for obtaining relative arrival times. Two proof-of-concept experiments were performed and some results of data obtained at Lake Travis and the Gulf of Mexico are presented     

A Colliding-Winds Interpretation of the U Orbital Variation of the Symbiotic Binary AG Pegasi

The photometric U orbital variations of the symbiotic binary AG Peg, caused by occultation of a bright gaseous region by the M giant in this system, are used to determine the geometry of this region whose existence is explained in the framework of the colliding winds model. The emission of the bright region varies and a particular value of its mean density of 2–3 10¹⁰ cm^-3 has been derived. The causes for the irregular variations of the light in the orbital minimum and maximum are discussed. It is also obtained, when the mass-loss rate of the hot companion of the system decreases to about 5 10^-10 Ms _⊙ yr^-1, this bright region will disappear. This revised version was published online in July 2006 with corrections to the Cover Date.     

Accretion disk spectra of super-luminal jet sources and ultra-luminous compact X-ray sources in nearby spiral galaxies

The Ultra-luminous Compact X-ray Sources (ULXs)in nearby spiral galaxies and the Galactic super-luminaljet sources sharethe common spectral characteristic that they haveextremely high disk temperatures which cannot be explainedin the framework of the standard accretion disk modelin the Schwarzschild metric. We have calculated an extreme Kerr disk model to examine if the Kerr disk model can instead explain the observed `too hot' accretion disk spectra.We found that the Kerr disk spectrum becomes significantly hardercompared to the Schwarzschild disk only when the disk is highlyinclined.For super-luminal jet sources, which are known to beinclined systems, the Kerr disk model may thuswork if we choose proper values for the black hole angular momentum. For the ULXs, however, the Kerr disk interpretation will be problematic,as is is highly unlikely that their accretion disks are preferentiallyinclined.     

Brightness Variation and Proper Motions of Three Objects in the Cygnus Region

The results of an investigation of three peculiar objects in the Cygnus region are given. One of them was already known to be a Herbig-Haro object. In accordance with its observational characteristics, the second object is, in all probability, a T Tauri star. The physical character of the last object is not yet entirely clear, but from the properties of its emission in the near infrared range it is probably also a T Tauri star. The proper motions of the objects have been measured. It is shown that all three objects are members of a tight system of the Trapezium type and evidently are the results of the successive fragmentation of an original protostellar body.     

The SuperCOSMOS Sky Survey – III. Astrometry

In this, the third in a series of three papers concerning the SuperCOSMOS Sky Survey, we describe the astrometric properties of the data base. We describe the algorithms employed in the derivation of the astrometric parameters of the data, and demonstrate their accuracies by comparison with external data sets using the first release of data, the South Galactic Cap survey. We show that the celestial coordinates, which are tied to the International Celestial Reference Frame via the Tycho–2 reference catalogue, are accurate to better than ±0.2 arcsec at J , R ∼19,18 , rising to ±0.3 arcsec at J , R ∼22,21 , with positional-dependent systematic effects from bright to faint magnitudes at the ∼0.1-arcsec level. The proper motion measurements are shown to be accurate to typically ±10 mas yr⁻¹ at J , R ∼19,18 , rising to ±50 mas yr⁻¹ at J , R ∼22,21 , and are tied to zero using the extragalactic reference frame. We show that the zero-point errors in the proper motions are ≤1 mas yr⁻¹ for R >17 , and are no larger than ∼10 mas yr⁻¹ for R <17 mas yr⁻¹ .     

On the temperature measurement from the O IV emission lines

The new calculations of the Oiv temperature-sensitive EUV line ratios are presented and compared with previous results.