Michael C. Storrie-Lombardi

Automated morphological classification of APM galaxies by supervised artificial neural networks

We train Artificial Neural Networks to classify galaxies based solely on the morphology of the galaxy images as they appear on blue survey plates. The images are reduced and morphological features such as bulge size and the number of arms are extracted, all in a fully automated manner. The galaxy sample was first classified by 6 independent experts. We use several definitions for the mean type of each galaxy, based on those classifications. We then train and test the network on these features. We find that the rms error of the network classifications, as compared with the mean types of the expert classifications, is 1.8 Revised Hubble Types. This is comparable to the overall rms dispersion between the experts. This result is robust and almost completely independent of the network architecture used.

Galaxies, Human Eyes, and Artificial Neural Networks

The quantitative morphological classification of galaxies is important for understanding the origin of type frequency and correlations with environment. However, galaxy morphological classification is still mainly done visually by dedicated individuals, in the spirit of Hubbles original scheme and its modifications. The rapid increase in data on galaxy images at low and high redshift calls for a re-examination of the classification schemes and for automatic methods. Here are shown results from a systematic comparison of the dispersion among human experts classifying a uniformly selected sample of more than 800 digitized galaxy images. These galaxy images were then classified by six of the authors independently. The human classifications are compared with each other and with an automatic classification by an artificial neural network, which replicates the classification by a human expert to the same degree of agreement as that between two human experts.

A Comparative study of morphological classifications of APM galaxies

We investigate the consistency of visual morphological classifications of galaxies by comparing classifications for 831 galaxies from six independent observers. The galaxies were classified on laser print copy images or on computer screen produced from scans with the Automated Plate Measuring (APM) machine. Classifications are compared using the Revised Hubble numerical type index T. We find that individual observers agree with one another with rms combined dispersions of between 1.3 and 2.3 type units, typically about 1.8 units. The dispersions tend to decrease slightly with increasing angular diameter and, in some cases, with increasing axial ratio

Neural computation as a tool for galaxy classification : methods and examples

(b/a)

Spectral classification with principal component analysis and artificial neural networks

. The agreement between independent observers is reasonably good but the scatter is non-negligible. In spite of the scatter the Revised Hubble T system can be used to train an automated galaxy classifier, e.g. an Artificial Neural Network, to handle the large number of galaxy images that are being compiled in the APM and other surveys.

Automated classification of stellar spectra – I. Initial results with artificial neural networks

We apply and compare various artificial neural network (ANN) and other algorithms for the automated morphological classification of galaxies. The ANNs are presented here mathematically, as non-linear extensions of conventional statistical methods in astronomy. The methods are illustrated using a selection of subsets from the ESO-LV catalogue, for which both machine parameters and human classifications are available. The main methods we explore are: (i) principal component analysis (PCA), which provides information on how independent and informative the input parameters are; (ii) encoder neural networks, which allow us to find both linear (PCA-like) and non-linear combinations of the input, illustrating an example of an unsupervised ANN; and (iii) supervised ANNs (using the backpropagation or quasi-Newton algorithm) based on a training set for which the human classification is known. Here the output for previously unclassified galaxies can be interpreted as either a continuous (analogue) output (for example T-type) or a Bayesian a posteriori probability for each class. Although the ESO-LV parameters are suboptimal, the success of the ANN in reproducing the human classification is 2 T-type units, similar to the degree of agreement between two human experts who classify the same galaxy images on plate material. We also examine the aspects of ANN configurations, reproducibility, scaling of input parameters and redshift information.

Morphological classication of galaxies by articial neural networks

Derived from non-linear signal processing strategies common to biological systems, neural network algorithms generalise classical data analysis techniques, e.g. Fourier analysis, Wiener filtering, and vector clustering algorithms. Conversely, multifactor analysis tools such as principal component analysis can function in a manner analogous to that of an unsupervised neural network. We have explored the use of principal component analysis for data pre-processing prior to classification of stellar spectra with a non-linear neural network. The strategy significantly enhances classification replicability, network stability, and convergence.