Probabilistic Interpretation of Feedforward Classification Network Outputs, with Relationships to Statistical Pattern Recognition

Bridle, John S.

doi:10.1007/978-3-642-76153-9_28

Cited by 873 publications

(499 citation statements)

References 1 publication

Supporting

Mentioning

467

Contrasting

Unclassified

Order By: Relevance

“…Obviously, binary data for species occurrence did not require normalization. Sigmoid activation functions were used in the hidden layer nodes [f(x)=1/(1−e − x )], whereas softmax activation functions (Bridle, 1990) were used in the output layer nodes. The softmax function scales the neural network outputs so that their sum is 1 and that each output can be regarded as a probability, i.e., in this particular case, as a membership value for each class of ecological status.…”

Section: Methodsmentioning

confidence: 99%

An expert system based on fish assemblages for evaluating the ecological quality of streams and rivers

Scardi

Cataudella

Dato

et al. 2008

Ecological Informatics

View full text Add to dashboard Cite

Fish species have been often used as indicators of environmental quality in aquatic ecosystems, while biotic indices based on fish have become common tools in ecological monitoring. Nevertheless, such indices are far from perfect, mainly because they are based on assumptions that sometimes are not met and because they cannot be optimized from a computational point of view. As any other method, they rely upon expert judgments for selecting relevant metrics, combining metrics into a score and defining thresholds between ecological status classes in the scoring scale. Provided that no procedure can be entirely objective in evaluating ecological status, as this very concept is inherently subjective, we propose a novel approach in which the unavoidable subjective elements only play a role in the earliest steps, while the subsequent optimization of the evaluation procedure is as objective as possible. An expert system, designed after this concept for Latium (Central Italy) river basins and based on a multilayer perceptron neural network, was developed and implemented into a Graphical User Interface (GUI) in order to make it easily accessible to non-technical users. The neural network reconstructs experts' judgments on the basis of a set of abiotic descriptors and fish assemblage composition, thus providing consensus estimates of ecological status for any river stretch. This approach allows easily the incorporation into the expert system of new data and new expert judgments as soon as they become available. However, the very first version of the expert system is already able to correctly classify 2 out of 3 cases, while the worst classification error does not exceed a single class of ecological status. (c) 2007 Elsevier B.V. All rights reserved

show abstract

Section: Methodsmentioning

confidence: 99%

An expert system based on fish assemblages for evaluating the ecological quality of streams and rivers

Scardi

Cataudella

Dato

et al. 2008

Ecological Informatics

View full text Add to dashboard Cite

show abstract

“…The genesis and renaissance of ANNs took place within various communities, and papers published during these periods reflected the disciplines involved: biology and cognition; statistical physics; and computer science. But it was not until the early 1990s that a probability-theoretic perspective emerged, which regarded ANNs as being within the framework of statistics [17,18,19,20].…”

Section: Multilayer Perceptronsmentioning

confidence: 99%

Perspectives and Prospects

Mareschal¹

2016

Introduction to Drug Disposition and Pharmacokinetics

View full text Add to dashboard Cite

In 1998, over 400 papers on artificial neural networks (ANNs) were published in the context of medicine, but why is there this interest in ANNs? And how do ANNs compare with traditional statistical methods? We propose some answers to these questions, and go on to consider the 'black box' issue. Finally, we briefly look at two directions in which ANNs are likely to develop, namely the use of Bayesian statistics and knowledgedata fusion.

show abstract

“…The network contained 43,257 trainable weights in total. As is standard for classification tasks, the softmax activation function was used at the output layer, with the cross-entropy objective function [1]. The network was trained using online gradient descent (weight updates after every training sequence) with a learning rate of 10 −6 and a momentum of 0.9.…”

Section: Air Freight Datamentioning

confidence: 99%

Multi-dimensional Recurrent Neural Networks

Graves

Fernández

Schmidhuber

2007

Lecture Notes in Computer Science

239

180

View full text Add to dashboard Cite

Abstract. Recurrent neural networks (RNNs) have proved effective at one dimensional sequence learning tasks, such as speech and online handwriting recognition. Some of the properties that make RNNs suitable for such tasks, for example robustness to input warping, and the ability to access contextual information, are also desirable in multi-dimensional domains. However, there has so far been no direct way of applying RNNs to data with more than one spatio-temporal dimension. This paper introduces multi-dimensional recurrent neural networks, thereby extending the potential applicability of RNNs to vision, video processing, medical imaging and many other areas, while avoiding the scaling problems that have plagued other multi-dimensional models. Experimental results are provided for two image segmentation tasks.

show abstract

Probabilistic Interpretation of Feedforward Classification Network Outputs, with Relationships to Statistical Pattern Recognition

Cited by 873 publications

References 1 publication

An expert system based on fish assemblages for evaluating the ecological quality of streams and rivers

An expert system based on fish assemblages for evaluating the ecological quality of streams and rivers

Perspectives and Prospects

Multi-dimensional Recurrent Neural Networks

Contact Info

Product

Resources

About