No abstract

Concerns the problem of finding weights for feed-forward networks in which threshold functions replace the more common logistic node output function. The advantage of such weights is that the complexity of the hardware implementation of such networks is greatly reduced. If the task to be learned does not change over time, it may be sufficient to find the correct weights for a threshold function network off-line and to transfer these weights to the hardware implementation. This paper provides a mathematical foundation for training a network with standard logistic function nodes and gradually altering the function to allow a mapping to a threshold unit network. The procedure is analogous to taking the limit of the logistic function as the gain parameter goes to infinity. It is demonstrated that, if the error in a trained network is small, a small change in the gain parameter will cause a small change in the network error. The result is that a network that must be implemented with threshold functions can first be trained using a traditional back propagation network using gradient descent, and further trained with progressively steeper logistic functions. In theory, this process could require many repetitions. In simulations, however, the weights have be successfully mapped to a true threshold network after a modest number of slope changes. It is important to emphasize that this method is only applicable to situations for which off-line learning is appropriate.

Six advanced very high resolution radiometer local area coverage arctic scenes are classified into 10 classes. These include water, solid sea ice, broken sea ice, snow-covered mountains, land, stratus over ice, stratus over water, cirrus over ice, cumulus over water, and multilayer cloudiness. Eight spectral and textural features are computed. The textural features are based upon the gray level difference vector method. Six different artificial intelligence classifiers are examined: (1) the feed forward back propagation neural network; (2) the probabilistic neural network; (3) the hybrid back propagation neural network; (4) the "don• care" perceptton network; (5) the "don• care" back propagation neural network; and (6) a fuzzy logic-based expert system. Accuracies in excess of 95% are obtained for all but the hybrid neural network. The "don• care" back propagation neural network produces the highest accuracies and also has low CPU requirements. Thin fog/stratus over ice is the class consistently with the lowest accuracy, otten misclassified as broken sea ice. Water, land, cirrus over ice, and snow-covered mountains are all classified with high accuracy (>98%). The high accuracy achieved in the present study can be traced to (1) accurate classifiers; (2) an excellent choice for the feature vector; and (3) accurate labeling. A sophisticated new interactive visual image classification system is used for the labeling. 1. INTRODUCTION With the growing awareness and debate over the potential changes associated with global climate change, the polar regions are receiving increased attention. Since greenhouse forcings are expected to be amplified in polar regions [WetheraM and Manabe, 1986; Schlesinger and Mitchell, 1987; Steffen and Lewis, 1988], these regions may act as early warning indicators of actual climate shifts. Global cloud distributions can be expected to be altered with increased greenhouse forcings. In the polar regions, cloud cover changes can be expected to have a significant effect on sea ice conditions [Shine and Crane, 1984] and on regional ice-albedo feedback [Barry et al., 1984]. In particular, polar stratus is very important to the polar heat balance and directly affects surface melting [Parkinson et al., 1987]. However, in order to monitor changes in polar surface conditions and polar cloudiness, more accurate procedures must be developed to distinguish between cloud and snow-covered surfaces.Owing to the similarity of cloud and snow-ice spectral signatures in both visible and infrared wavelengths, it is difficult to distinguish clouds from surface features in the polar regions. In the visible channels, thin ice, ice fragments, wet ice, and pancake ice have low albedoes and can be misinterpreted as water, melt ponds, or as thin cloud/haze. Persistent surface inversions and low clouds in winter and near isothermal structure and extensive stratiform clouds in summer limit discrimination in the in•rrared

Abstract. The accuracy and efficiency of four approaches to identifying clouds and aerosols in remote sensing imagery are compared. These approaches are as follows: a maximum likelihood classifier, a paired histogram technique, a hybrid class elimination approach, and a backpropagation neural network. Regional comparisons were conducted on advanced very high resolution radiometer (AVHRR) local area coverage (LAC) scenes from the polar regions, desert areas, and regions of biomass-burning, areas which are known to be particularly difficult. For the polar, desert, and biomass burning regions, the maximum likelihood classifier achieved 94-97% accuracy, the neural network achieved 95-96% accuracy, and the paired histogram approach achieved 93-94% accuracy. The primary advantage to the class elimination scheme lies in its speed; its accuracy of 94-96% is comparable to that of the maximum likelihood classifier. Experiments also clearly demonstrate the effectiveness of decomposing a single global classifier into separate regional classifiers, since the regional classifiers can be more finely tuned to recognize local conditions. In addition, the effectiveness of using composite features is compared to the simpler approach of using the five AVHRR channels and the reflectance of channel 3 treated as a sixth channel as the elements of the feature vector. The results varied, demonstrating that the features cannot be chosen independently of the classifier to be used. It is also shown that superior results can obtained by training the classifiers using subclass information and collapsing the subclasses after classification. Finally, ancillary data were incorporated into the classifiers, consisting of a land/water mask, a terrain map, and a computed sunglint probability. While the neural network did not benefit from this information, the accuracy of the maximum likelihood classifier improved by 1%, and the accuracy of the paired histogram method increased by up to 4%.

We describe a massively parallel implementation of genetic algorithms using a MasPar MP-1 data-parallel computer. Modification of the sequential genetic algorithm required that several important issues be addressed, in particular how to implement the selection operator with a minimum of inter-processor communication. A speed up of at least 145 times relative to a comparable sequential algorithm running on a Sun SparcStation was achieved. The performance of the MasPar is even more impressive when populations are too large to tit into the physical memory of the Spare.