Search

We develop and demonstrate a probabilistic method for classifying rare objects in surveys with the particular goal of building very pure samples. It works by modifying the output probabilities from a classifier so as to accommodate our expectation (priors) concerning the relative frequencies of different classes of objects. We demonstrate our method using the Discrete Source Classifier (DSC), a supervised classifier currently based on support vector machines, which we are developing in preparation for the Gaia data analysis. DSC classifies objects using their very low resolution optical spectra. We look in detail at the problem of quasar classification, because identification of a pure quasar sample is necessary to define the Gaia astrometric reference frame. By varying a posterior probability threshold in DSC, we can trade off sample completeness and contamination. We show, using our simulated data, that it is possible to achieve a pure sample of quasars (upper limit on contamination of 1 in 40 000) with a completeness of 65 per cent at magnitudes of , and 50 per cent at , even when quasars have a frequency of only 1 in every 2000 objects. The star sample completeness is simultaneously 99 per cent with a contamination of 0.7 per cent. Including parallax and proper motion in the classifier barely changes the results. We further show that not accounting for class priors in the target population leads to serious misclassifications and poor predictions for sample completeness and contamination. We discuss how a classification model prior may, or may not, be influenced by the class distribution in the training data. Our method controls this prior and so allows a single model to be applied to any target population without having to tune the training data and retrain the model. [ABSTRACT FROM AUTHOR]

We aim to obtain a complete sample of redshift radio quasi-stellar objects (QSOs) from the Faint Images of the Radio Sky at Twenty cm survey (FIRST) sources having star-like counterparts in the Sloan Digital Sky Survey (SDSS) Data Release 5 (DR5) photometric survey . Our starting sample of 8665 FIRST–DR5 pairs includes 4250 objects with spectra in DR5, 52 of these being QSOs. We found that simple supervised neural networks, trained on the sources with DR5 spectra, and using optical photometry and radio data, are very effective for identifying high- z QSOs in a sample without spectra. For the sources with DR5 spectra the technique yields a completeness (fraction of actual high- z QSOs classified as such by the neural network) of 96 per cent, and an efficiency (fraction of objects selected by the neural network as high- z QSOs that actually are high- z QSOs) of 62 per cent. Applying the trained networks to the 4415 sources without DR5 spectra we found QSO candidates. We obtained spectra of 27 of them, and 17 are confirmed as high- z QSOs. Spectra of 13 additional candidates from the literature and from SDSS Data Release 6 (DR6) revealed seven more QSOs, giving an overall efficiency of 60 per cent (24/40). None of the non-candidates with spectra from NASA/IPAC Extragalactic Database (NED) or DR6 is a QSO, consistently with a high completeness. The initial sample of high- z QSOs is increased from 52 to 76 sources, i.e. by a factor of 1.46. From the new identifications and candidates we estimate an incompleteness of SDSS for the spectroscopic classification of FIRST QSOs of 15 per cent for . [ABSTRACT FROM AUTHOR]

We compare the performance of two automated classification algorithms, k-dimensional tree (kd-tree) and support vector machines (SVMs), to separate quasars from stars in the data bases of the Sloan Digital Sky Survey (SDSS) and the Two-Micron All Sky Survey (2MASS) catalogues. The two algorithms are trained on subsets of SDSS and 2MASS objects whose nature is known via spectroscopy. We choose different attribute combination as input patterns to train the classifier using photometric data only and present the classification results obtained by these two methods. Performance metrics, such as precision and recall, true positive rate and true negative rate, F-measure, G-mean and Weighted Accuracy, are computed to evaluate the performance of the two algorithms. The study shows that both kd-tree and SVMs are effective automated algorithms to classify point sources. SVMs show slightly higher accuracy, but kd-tree requires less computation time. Given different input patterns based on various parameters (e.g. magnitudes, colour information), we conclude that both kd-tree and SVMs show better performance with fewer features. What is more, our results also indicate that the accuracy using the four colours ( u− g, g− r, r− i and i− z) and r magnitude based on SDSS model magnitudes adds up to the highest value. The classifiers trained by kd-tree and SVMs can be used to solve the automated classification problems faced by the virtual observatory (VO); moreover, they can all be applied for the photometric preselection of quasar candidates for large survey projects in order to optimise the efficiency of telescopes. [ABSTRACT FROM AUTHOR]