The effect of training set size and composition on artificial neural network classification

Foody, Giles M.; Mcculloch, M. B.; Yates, William B.

doi:10.1080/01431169508954507

Cited by 165 publications

(84 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…As expected, the accuracies increased as more training data were used due to an increased likelihood of capturing spectral and temporal differences within and among percent tree levels [15][16][17]52]. The accuracy differences were negligible (≤0.07% RMSE) when the most training data (8% per tree) were used.…”

Section: Discussionsupporting

confidence: 58%

Demonstration of Percent Tree Cover Mapping Using Landsat Analysis Ready Data (ARD) and Sensitivity with Respect to Landsat ARD Processing Level

et al. 2018

View full text Add to dashboard Cite

Abstract:The recently available Landsat Analysis Ready Data (ARD) are provided as top of atmosphere (TOA) and atmospherically corrected (surface) reflectance tiled products and are designed to make the U.S. Landsat archive for the United States straightforward to use. In this study, the utility of ARD for 30 m percent tree cover mapping is demonstrated and the impact of different ARD processing levels on mapping accuracy examined. Five years of Landsat 5 and 7 ARD over 12 tiles encompassing Washington State are considered using an established bagged regression tree methodology and training data derived from Goddard LiDAR Hyperspectral & Thermal Imager (G-LiHT) data. Sensitivity to the amount of training data is examined with increasing mapping accuracy observed as more training data are used. Four processing levels of ARD are considered independently and the mapped results are compared: (i) TOA ARD; (ii) surface ARD; (iii) bidirectional reflectance distribution function (BRDF) adjusted atmospherically corrected ARD; and (iv) weekly composited BRDF adjusted atmospherically corrected ARD. The atmospherically corrected ARD provide marginally the highest mapping accuracies, although accuracy differences are negligible among the four (≤0.07% RMSE) when modest amounts of training data are used. The TOA ARD provide the most accurate maps compared to the other input data when only small amounts of training data are used, and the least accurate maps otherwise. The results are illustrated and the implications discussed.

show abstract

Section: Discussionsupporting

confidence: 58%

Demonstration of Percent Tree Cover Mapping Using Landsat Analysis Ready Data (ARD) and Sensitivity with Respect to Landsat ARD Processing Level

et al. 2018

View full text Add to dashboard Cite

show abstract

“…The network's architecture and algorithm parameters were defined from an evaluation of several hundreds of candidate networks. Previous studies have shown that the training set, notably in terms of its size and composition can have a marked impact on the accuracy of classification by a neural network [34][35][36]. Moreover, it is apparent that the individual training cases vary in importance, with those lying close to the class borders most informative and helpful in determining the location of the classification hyperplanes [37].…”

Section: Data and Methods Of Classificationmentioning

confidence: 99%

A relative evaluation of multiclass image classification by support vector machines

Foody

Mathur

2004

IEEE Trans. Geosci. Remote Sensing

Self Cite

872

394

View full text Add to dashboard Cite

Support vector machines (SVM) have considerable potential as classifiers of remotely sensed data. A constraint on their application in remote sensing has been their binary nature, requiring multi-class classifications to be based upon a large number of binary analyses. Here, an approach for multi-class classification of airborne sensor data by a single SVM analysis is evaluated against a series of classifiers that are widely used in remote sensing, with particular regard to the effect of training set size on classification accuracy. In addition to the SVM, the same data sets were classified using a discriminant analysis, decision tree and multilayer perceptron neural network. The accuracy statements of the classifications derived from the different classifiers were compared in a statistically rigorous fashion that accommodated for the related nature of the samples used in the analyses. For each classification technique, accuracy was positively related with the size of the training set. In general, the most accurate classifications were derived from the SVM approach, and with the largest training set the SVM classification was significantly (p<0.05) more accurate (93.75%) than that derived from the discriminant analysis (90.00%) and decision tree algorithms (90.31%). Although each classifier could yield a very accurate classification, >90% correct, the classifiers differed in the ability to correctly label individual cases and so may be suitable candidates for an ensemble based approach to classification.3

show abstract

“…The MLP neural network-a supervised model that uses single or multilayer perceptrons to approximate the inherent input-out-put relationships-is the most commonly used network model for image classification in remote sensing [26][27][28]. MLP networks are typically trained with the supervised backpropagation (BP) algorithm [23] and consist of one input layer, one or more hidden layers, and one output layer ( Figure 1).…”

Section: Neural Network Classification Approachesmentioning

confidence: 99%

“…Both [29] and [30] found that an enhanced neural network can be achieved by incorporating a momentum term (the past increment to the weight) to speed up and stabilize the BP learning. Although there are many examples of successful MLP applications [11,17,[26][27][28]31,32], it is widely recognized that MLPs are sensitive to many operational factors including the size and quality of the training data set, network architecture, training parameters, and over-fitting problems. These factors are application-dependent and best addressed on a case-by-case basis.…”

Section: Neural Network Classification Approachesmentioning

confidence: 99%

An Automated Artificial Neural Network System for Land Use/Land Cover Classification from Landsat TM Imagery

Yuan¹,

2009

View full text Add to dashboard Cite

This paper focuses on an automated ANN classification system consisting of two modules: an unsupervised Kohonen's Self-Organizing Mapping (SOM) neural network module, and a supervised Multilayer Perceptron (MLP) neural network module using the Backpropagation (BP) training algorithm. Two training algorithms were provided for the SOM network module: the standard SOM, and a refined SOM learning algorithm which incorporated Simulated Annealing (SA). The ability of our automated ANN system to perform Land-Use/Land-Cover (LU/LC) classifications of a Landsat Thematic Mapper (TM) image was tested using a supervised MLP network, an unsupervised SOM network, unsupervised SOM networks. Additionally, the classification performance of the refined SOM network was found to be significantly better than that of the standard SOM network essentially due to the incorporation of SA. This is mainly due to the SA-assisted classification utilizing the scheduling cooling scheme. It is concluded that our automated ANN classification system can be utilized for LU/LC applications and will be particularly useful when traditional statistical classification methods are not suitable due to a statistically abnormal distribution of the input data.

show abstract

The effect of training set size and composition on artificial neural network classification

Cited by 165 publications

References 15 publications

Demonstration of Percent Tree Cover Mapping Using Landsat Analysis Ready Data (ARD) and Sensitivity with Respect to Landsat ARD Processing Level

Demonstration of Percent Tree Cover Mapping Using Landsat Analysis Ready Data (ARD) and Sensitivity with Respect to Landsat ARD Processing Level

A relative evaluation of multiclass image classification by support vector machines

An Automated Artificial Neural Network System for Land Use/Land Cover Classification from Landsat TM Imagery

Contact Info

Product

Resources

About