A detailed comparison of backpropagation neural network and maximum-likelihood classifiers for urban land use classification

Paola, J.D.; Schowengerdt, Robert A.

doi:10.1109/36.406684

Cited by 342 publications

(149 citation statements)

References 16 publications

Supporting

Mentioning

142

Contrasting

Unclassified

Order By: Relevance

“…To this end, we evaluate five prominent classifiers proposed in [25], [23], [14], [29], [9], [24]. These classifiers are quite diverse.…”

Section: Accuracy Evaluation Of Existing Techniquesmentioning

confidence: 99%

“…These classifiers are quite diverse. Naive Bayes [25] is a probabilistic classifier; while Back Propagation Neural Network (BPNN) [23] and Radial Basis Function Network (RBFN) [14] belong to the category of neural networks. In comparison, Kstar [9] is a statistical classifier and J48 [24] is a decision tree classifier.…”

Section: Accuracy Evaluation Of Existing Techniquesmentioning

confidence: 99%

See 1 more Smart Citation

Keystroke-Based User Identification on Smart Phones

Zahid

Shahzad

Khayam

et al. 2009

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Smart phones are now being used to store users' identities and sensitive information/data. Therefore, it is important to authenticate legitimate users of a smart phone and to block imposters. In this paper, we demonstrate that keystroke dynamics of a smart phone user can be translated into a viable feature set for accurate user identification. To this end, we collect and analyze keystroke data of 25 diverse smart phone users. Based on this analysis, we select six distinguishing keystroke features that can be used for user identification. We show that these keystroke features for different users are diffused and therefore a fuzzy classifier is well-suited to cluster and classify them. We then optimize the front-end fuzzy classifier using Particle Swarm Optimizer (PSO) and Genetic Algorithm (GA) as back-end dynamic optimizers to adapt to variations in usage patterns. Finally, we provide a novel keystroke dynamics based PIN verification mode to ensure information security on smart phones. The results of our experiments show that the proposed user identification system has an average error rate of 2% after the detection mode and the error rate of rejecting legitimate users is dropped to zero after the PIN verification mode. We also compare error rates (in terms of detecting both legitimate users and imposters) of our proposed classifier with 5 existing state-of-the-art techniques for user identification on desktop computers. Our results show that the proposed technique consistently and considerably outperforms existing schemes.

show abstract

“…To this end, we evaluate five prominent classifiers proposed in [25], [23], [14], [29], [9], [24]. These classifiers are quite diverse.…”

Section: Accuracy Evaluation Of Existing Techniquesmentioning

confidence: 99%

Section: Accuracy Evaluation Of Existing Techniquesmentioning

confidence: 99%

Keystroke-Based User Identification on Smart Phones

Zahid

Shahzad

Khayam

et al. 2009

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…Apart from that, extraction of class-conditional spectral parameters using mean and standard deviation from supervised training sites of pure pixels used in the FMLP is dependent on the distribution of the reflectance values. In [17] it is observed that Neural Network classifiers as compared to statistical classifiers are nonparametric (distribution free). Statistical classifiers give incorrect results when reflectance values of classes are very close.…”

Section: Proposed Methodsmentioning

confidence: 99%

Nonparametric Neural Network Model Based on Rough-Fuzzy Membership Function for Classification of Remotely Sensed Images

Kumar

Agrawal

2006

Computer Vision, Graphics and Image Processing

View full text Add to dashboard Cite

Abstract.A nonparametric neural network model based on Rough-Fuzzy Membership function, multilayer perceptron, and back-propagation algorithm is described. The described model is capable to deal with rough uncertainty as well as fuzzy uncertainty associated with classification of remotely sensed multi-spectral images. The input vector consists of membership values to linguistic properties while the output vector is defined in terms of rough fuzzy class membership values. This allows efficient modeling of indiscernibility and fuzziness between patterns by appropriate weights being assigned to the backpropagated errors depending upon the Rough-Fuzzy Membership values at the corresponding outputs. The effectiveness of the model is demonstrated on classification problem of IRS-P6 LISS IV images of Allahabad area. The results are compared with statistical (Minimum Distance), conventional MLP, and FMLP models.

show abstract

“…Therefore, the initial learning rate and momentum are not crucial to the success of training stage. Also training speed is increased because the learning rate is adjusted to the highest value that does not cause instability (Paola and Schowengerdt, 1997). For accuracy assessment, a binary image is extracted from image-truth, assigning 1 to road and 0 to background pixels.…”

Section: Methodologiesmentioning

confidence: 99%

Road detection from high-resolution satellite images using artificial neural networks

Mokhtarzade

Zoej

2007

International Journal of Applied Earth Observation and Geoinfor

162

View full text Add to dashboard Cite

ABSTRACT:In this article, the possibility of using artificial neural networks for road detection from high resolution satellite images is tested on a part of RGB Ikonos and Quick-Bird images from Kish Island and Bushehr Harbour respectively. Then, the effects of different input parameters on network's ability are verified to find out optimum input vector for this problem. A variety of network structures with different iteration times are used to determine the best network structure and termination condition in training stage. It was discovered when the input parameters are made up of spectral information and distances of pixels to road mean vector in a 3*3 window, network's ability in both road and background detection can be improved in comparison with simple networks that just use spectral information of a single pixel in their input vector.

show abstract

A detailed comparison of backpropagation neural network and maximum-likelihood classifiers for urban land use classification

Cited by 342 publications

References 16 publications

Keystroke-Based User Identification on Smart Phones

Keystroke-Based User Identification on Smart Phones

Nonparametric Neural Network Model Based on Rough-Fuzzy Membership Function for Classification of Remotely Sensed Images

Road detection from high-resolution satellite images using artificial neural networks

Contact Info

Product

Resources

About