Face Detection Using CMAC Neural Network

Fashandi, Homa; Moin, Mohammad Shahram

doi:10.1007/978-3-540-24844-6_111

Cited by 2 publications

(2 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…8 Originally, the CMAC was proposed as a function modeler for robotic controllers, 7 but has been extensively used in reinforcement learning 9,10 and also as a classifier. [11][12][13][14] The training method proposed by Albus was an iterative algorithm based on global error minimization. Empirical evidence, presented in this paper, suggests that this algorithm has linear time complexity.…”

Section: Cerebellar Model Articulation Controllermentioning

confidence: 99%

“…This is sufficient for two class problems, and is the most often cited in the literature. 12,13,22,23 For problems with more than two classes, one could define threshold values such as to divide the scalar range of z into the number of classes to be represented: (2) represents a scalar mapping. Using this mapping, the CMAC can be used as a classifier if, during the training phase, weights are adjusted to make the output z approach a suitable target value.…”

Section: Output Mappingmentioning

confidence: 99%

See 1 more Smart Citation

A Cerebellar Model Classifier for Data Mining With Linear Time Complexity

Cornforth

2006

Int. J. Comp. Intel. Appl.

View full text Add to dashboard Cite

Techniques for automated classification need to be efficient when applied to large datasets. Machine learning techniques such as neural networks have been successfully applied to this class of problem, but training times can blow out as the size of the database increases. Some of the desirable features of classification algorithms for large databases are linear time complexity, training with only a single pass of the data, and accountability for class assignment decisions. A new training algorithm for classifiers based on the Cerebellar Model Articulation Controller (CMAC) possesses these features. An empirical investigation of this algorithm has found it to be superior to the traditional CMAC training algorithm, both in accuracy and time required to learn mappings between input vectors and class labels. . 299 Int. J. Comp. Intel. Appl. 2006.06:299-313. Downloaded from www.worldscientific.com by NANYANG TECHNOLOGICAL UNIVERSITY on 08/26/15. For personal use only.300 D. CornforthMany algorithms for automated classification have an inherently non-linear relationship between time taken by the algorithm to run and the number of training examples. Analysis methods that work well for small data sets are completely impractical when applied to larger data sets. For example, training of a neural network using back-propagation is known to be NP-complete. 5 Some studies suggest that evolutionary algorithms have polynomial time complexity. 6 The work presented here investigates classification algorithms based on the Cerebellar Model Articulation Controller (CMAC), 7 which have linear time complexity.Global error minimization techniques, such as back-propagation, require multiple traversals of the data set during training. If the training set is very large, it cannot fit inside the memory of the machine. This will result in multiple disk read/write operations, which are relatively costly in time and can contribute greatly to data processing time. Current approaches include compression or summary of the data set before processing, and redesign of analysis tools so that analysis can be completed with only one pass of the data. This paper shows how the original CMAC training algorithm, which normally uses an iterative global error minimization technique, may be adapted so that the training set only needs to be accessed once.The usefulness of a classification algorithm may be enhanced by providing an explanation for each class assignment decision. This could take the form of a set of rules that contribute to the assignment, or a probability for each class, given the input. Black box methods such as neural networks do not naturally lend themselves to this form of analysis. The new algorithm described here provides accountability for class assignment decisions in the form of class probabilities.In this paper, I propose the Kernel Addition Training Algorithm (KATA) as a more effective learning algorithm for the CMAC when used as a classifier. The proposed method requires only a single pass of the data and provides a probability mode...

show abstract

Section: Cerebellar Model Articulation Controllermentioning

confidence: 99%

Section: Output Mappingmentioning

confidence: 99%

A Cerebellar Model Classifier for Data Mining With Linear Time Complexity

Cornforth

2006

Int. J. Comp. Intel. Appl.

View full text Add to dashboard Cite

show abstract

Face detection by color and multilayer feedforward neural network

Lin

2005 IEEE International Conference on Information Acquisition

View full text Add to dashboard Cite

In this paper, we introduce a novel approach for automatic detection of human faces embedded in dissimilar lighting. The proposed system consists of two primary parts. The first part is to convert the input RGB color images to a binary image directly using color segmentation. Because the absolute values of r, g, and b are totally different with the various skin colors in the altered lighting conditions and the relative value between r, g, and b are almost similar with the different skin colors in changed brightness circumstances, we use the relative value between r, g, and b in the color segmentation process to binarize the RGB color images directly instead of color images to gray level images, then binary ones . For this reason, our system is very robust for different lighting conditions. The second part of the proposed system is to search the potential face regions and perform the task of face detection. In the second part, each face candidate is obtained from the isosceles-triangle criterion that is based on the rules of "the combination of two eyes and one mouth", and then to be normalized to a standard size (60*60 pixels). Next, each of these normalized potential face regions are fed to neural networks function to obtain the location of the face region. The proposed face detection system can detect color multiple faces embedded in dissimilar lighting conditions. Moreover, it can conquer different size, varying pose and expression. Experimental results demonstrate that an approximately 97% success rate is achieved and the relative false estimation rate is very low.

show abstract

Face Detection Using CMAC Neural Network

Cited by 2 publications

References 10 publications

A Cerebellar Model Classifier for Data Mining With Linear Time Complexity

A Cerebellar Model Classifier for Data Mining With Linear Time Complexity

Face detection by color and multilayer feedforward neural network

Contact Info

Product

Resources

About