| prev toc next |
The availability for the first time of huge astronomical spectroscopic surveys such as the SDSS and 2Df, with more than { spectra each, will allow the accurate determination of intrinsic physical parameters of a large number of galaxies, including the age distribution and metallicity of their stellar populations.
The importance of the accurate knowledge of these parameters for cosmological studies and for the understanding of galaxy formation and evolution cannot be overestimated. Template fitting has been used to carry out estimates of the distribution of age and metallicity from spectral data. Although this technique achieves good results, it is very expensive in terms of computing time and therefore can be applied only to small samples.
Starting from a grid of theoretical population synthesis models we constructed a set of fully theoretical model galaxies with a distribution of ages, metallicities and intrinsic reddening. Using this set we have explored a new method that maximizes speed and accuracy. Our proposed technique combines standard least-squares fitting with an active instance-based machine learning algorithm. Experimental results show that this method can estimate with high speed and accuracy the physical parameters of the stellar populations. Based on empirical evidence we believe that this method can be applied with equal success to other astronomical problems, reducing the computational cost and thus providing the capability of analyzing larger quantities of astronomical data.
The 4th order cosine coefficient of Fourier harmonics, so called boxiness, is known to be one of the important parameters for automated classification of early type galaxies. So far, boxiness is calculated from isophote or brightness fitting, and it costs time and CPU. Inspired by Stokes parameters, I developed a method to estimate boxiness from moments. The algorithm and its accuracy will be presented.
We present QualityFITS, an element of the TERAPIX pipeline which allows a rapid quality assessment (background level, seeing, galaxy and star counts, etc...) on FITS images.
Among the many celestial objects in the universe, galaxies offer insights as to how the universe was formed and is continuing to develop. The morphological classification of galaxies is important just for this reason. The challenge lies in classifying the estimated billions of galaxies that are in the universe. The Hubble Deep Field has already produced images, which contain thousands of galaxies in a region of the sky, which is dark not only to the naked eye, but also to very large ground based telescopes. The Sloan Digital Sky Survey has started to produce vast amounts of data, which would take expert human classifiers a tremendous amount of time to sift through the images and classify the galaxies one by one. The automated procedure described here uses an image enhancing technique, segmentation, shape feature extraction and a supervised artificial neural network to classify the galaxies. When trained to classify galaxies as E/S0 or S, the network is able to learn 98.3% of the galaxies correctly and identify 89.9% of galaxy images in a test set. The major challenge is in the development of robust and automated segmentation schemes. With manually threshold images and Difference Boosting Neural Network we were able to achieve considerable success in developing a supervised classifier capable of sorting galaxies into subclasses.
This poster reviews a number of new routines in the NEMO package that can be used to fit particle distributions (e.g. from N-body simulations, but also grid based simulations) to observations. Non-linear fitting, especially those with many parameters, present themselves with many problems. Genetic programming is one such solution. In this poster I will present some galaxy dynamics fitting techniques, in particular those of velocity fields of individual galaxies and interacting galaxies, and apply new techniques such as genetic programming.
We present a system for the electronic administration of Astronomical image plates. This system is based on the digitization of the photographic plates to make the identification of spectral lines in low resolution spectra automatic, employing the techniques of image processing and image analysis. In the past, all the studies on photograhic plates to select objects by their color, broad emission lines, absorption lines, narrow-band and broad- band photometric indices, etc. was made by visual inspection of the plates. Today, we can make these studies automatically from a digitized image of plates . The astronomical plates used in this work were taken with the Tonantzintla Schmidt Camera 76.20/ 66.04/ 231.1 cm using the 3.96 degrees objective prism. Some algorithms assisted by computer already exist, those algorithms do not work automatically in the selection of objects and the identification of their spectral lines. We propose to make an identification of several spectral features in each spectrum. We also consider other factors as emulsion, objective prism and telescope characteristics in order to achieve this objective. Our goal is to obtain a system capable to obtain the most of the objects in an image and to give the user as a result the spectral characteristics of each them. We will apply this system to different objects with the purpose of comparing the results obtained by this method with those of the visual inspection technique.
In this work we present Evolution Strategies (ES) as an efficient method to approximate the initial conditions of the main interacting group of three galaxies in M81.
The M81 group is one of the nearest groups of galaxies. Its biggest galaxy, M81, sits in the core of the group together with its two nearby companions M82 and NGC3077. The interaction between these three galaxies is very well defined on an image taken in HI. In this first attempt we use non-self-gravitating simulations to approximate the initial conditions; even with that restriction our method reproduces the density distribution of the three galaxies with great precision.
Results presented here show that ES is an ideally suited method to work in optimization problems in Astrophysics, where the solution is hard to find by common methods. In particular we argue that ES is a good method to find initial conditions of groups of interacting galaxies, where a large number of parameters need to be determined.
Recent spectroscopic surveys of nearby AGN have proven that a large fraction show high-order hydrogen Balmer absorption lines in the near-UV. These features are characteristic of young stars and therefore represent a strong evidence of the presence of recent star formation in these galaxies. From a theoretical point of view, it is very important to determine the age of these starbursts, in order to understand the nature of the starburst-AGN connection and its relation with galaxy formation and evolution. The characterization of the nuclear star forming region (its age and mass) is very difficult to achieve in AGN, due to the contamination of the nuclear stellar absorption lines by the AGN component itself.
We present a new technique to determine the age of nuclear starbursts in galaxies with AGN using an ensemble of classifiers. The classifiers are specialized in the profile shape of high-order Balmer and Calcium K absorption lines and therefore very insensitive to the AGN contamination effect.
An ensemble of classifiers is a group of classifiers whose outputs are combined in some way, usually by voting. Ensembles of classifiers normally have better accuracy than the individual classifiers that make them up. For this work, each member of the ensemble was created using a randomly selected attribute set, and each applied the locally weighted regression algorithm, an instance based learning method that explicitly retains the training data and uses them to build a local linear model, valid only in the neighborhood of the point of interest, each time a prediction needs to be made. Our training data consist of 23 high spectral resolution synthetic models of starbursts of different ages. Each classifier was trained with randomly generated subsets of features, and the ensemble was tested using ten-fold cross-validation and an analytically classified test set. An accuracy of about 0.3dex in logarithmic age was achieved.
The method was then applied to the optical/near UV spectra of nuclear regions of nearby Seyfert galaxies covering the wavelength region 3600-5300 Angstroms and it was found to be very insensitive to the AGN dilution. The results obtained by means of machine learning are compared with those produced by exhaustive search in terms of time and precision. We conclude that the automatic learning method greatly surpasses the performance of the traditional method.
\abstract{ Two factors that are known to have direct influence on the classification accuracy of any neural network are (1) the network complexity and (2) the representational accuracy of the training data. While pruning algorithms are used to tackle the complexity problem, no direct solutions are known for the second. Selecting training data at random from the sample space is the most popular method followed. Despite its simplicity, this method does not ensure nor guarantee that the training would be optimal. In this brief paper, we present a new method that is specific to a \textit{difference boosting neural network} (DBNN) but could probably be extended to other networks as well. The method is iterative and fast, ensuring optimal selection of the minimum training data from a larger set in an automated manner. We test the performance of the new method on the some of the well known datasets from the UCI repository for benchmarking machine learning tools and show that the performance of the new method in almost all cases is better than that in any published method of comparable network complexity and that it requires only a fraction of the usual training data, thereby, making learning faster and more generic. }
| prev toc next |