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.     

Sources and sinks of sea salt in a newfoundland blanket bog

Over an oceanic peatland, the concentration of Na in fog averaged 38.1 mgl^?1 compared with 1.8 mgl^?1 in rain, resulting in a significant flux of mineral elements to the surface. Between 16 May and 20 June 1990 the average mass flux of Na to the bog surface by fog, rain, and dry deposition was 21.9, 10.4 and 7.0 mg m^?2 d^?1. There was little long-term storage of Na within the peatland system, where Na losses measured in stream runoff averaged 34.8 mg m² d^?1, and deep groundwater losses 4 mg m^?2 d^?1. Calcium and Mg were preferentially retained in the organic soil, whereas K was relatively mobile. Potassium tended to become concentrated in the unsaturated zone. Stream runoff had a consistently higher pH than groundwater, corresponding to higher Ca and Mg concentrations, which may have been from mineral sources in the headwater ponds. Otherwise, the stream water chemistry was closely related to groundwater in the upper layers of the peat deposit.     

Main stages of evolution of matter in the universe. I

The problem of the equation of state of cosmic matter is discussed and the constants of integration in the Friedmann solutions are determined. Translated from Astrofizika, Vol. 40, No. 1, pp. 117–124, January–March, 1997.     

A chum scale pattern analyzer and stock identification based onscale patterns by discriminant function and power spectrum analysis

An optoelectronic instrument has been developed to analyze the scales and scale patterns of chum salmon. This device, with specially developed software, has been used to study populations of salmon. Fish from two different rivers approximately 200 km apart from each other were grouped with an accuracy of 82% or more. Machine-measured ages coincided with those derived from visual estimation in approximately 80% of the fish sampled. These results suggest the proposed system could be of practical use for measuring and analyzing the scale patterns that allow one to accurately measure both natural and farmed populations of salmon     

Examples of clustering around high redshift radio-galaxies

Published in Astrofizika, Vol. 38, No. 4, pp. 677–680, October–December, 1995.     

Comparison of results from several AOGCMs for global and regional sea-level change 1900–2100

 Sea-level rise is an important aspect of climate change because of its impact on society and ecosystems. Here we present an intercomparison of results from ten coupled atmosphere-ocean general circulation models (AOGCMs) for sea-level changes simulated for the twentieth century and projected to occur during the twenty first century in experiments following scenario IS92a for greenhouse gases and sulphate aerosols. The model results suggest that the rate of sea-level rise due to thermal expansion of sea water has increased during the twentieth century, but the small set of tide gauges with long records might not be adequate to detect this acceleration. The rate of sea-level rise due to thermal expansion continues to increase throughout the twenty first century, and the projected total is consequently larger than in the twentieth century; for 1990–2090 it amounts to 0.20–0.37 m. This wide range results from systematic uncertainty in modelling of climate change and of heat uptake by the ocean. The AOGCMs agree that sea-level rise is expected to be geographically non-uniform, with some regions experiencing as much as twice the global average, and others practically zero, but they do not agree about the geographical pattern. The lack of agreement indicates that we cannot currently have confidence in projections of local sea-level changes, and reveals a need for detailed analysis and intercomparison in order to understand and reduce the disagreements. Received: 1 September 2000 / Accepted: 20 April 2001     

A new pattern for the north–south asymmetry of sunspots

We study the solar cycle evolution during the last 8 solar cycles using a vectorial sunspot area called the LA (longitudinal asymmetry) parameter. This is a useful measure of solar activity in which the stochastic, longitudinally evenly distributed sunspot activity is reduced and which therefore emphasizes the more systematic, longitudinally asymmetric sunspot activity. Interesting differences are found between the LA parameter and the more conventional sunspot activity indices like the (scalar) sunspot area and the sunspot number. E.g., cycle 19 is not the highest cycle according to LA. We have calculated the separate LA parameters for the northern and southern hemisphere and found a systematic dipolar-type oscillation in the dominating hemisphere during high solar activity times which is reproduced from cycle to cycle. We have analyzed this oscillation during cycles 16–22 by a superposed epoch method using the date of magnetic reversal in the southern hemisphere as the zero epoch time. According to our analysis, the oscillation starts by an excess of the northern LA value in the ascending phase of the solar cycle which lasts for about 2.3 years. Soon after the maximum northern dominance, the southern hemisphere starts dominating, reaching its minimum some 1.2–1.7 years later. The period of southern dominance lasts for about 1.6 years and ends, on an average, slightly before the end of magnetic reversal.