KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Settings

Author: search.filters.author.Jearanaitanakij, KietikulSubject: search.filters.subject.Artificial neural network

Search Results

Now showing 1 - 3 of 3
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Item,
    Spartan simplicity: A pruning algorithm for neural nets
    (2008-08-01)
    Jearanaitanakij, Kietikul
    ;
    Pinngern, Ouen
    Having more hidden units than necessary can produce a neural network that has a poor generalization. This paper proposes a new algorithm for pruning unnecessary hidden units away from the single-hidden layer feedforward neural networks, resulting in a Spartan network. Our approach is simple and easy to implement, yet produces a very good result. The idea is to train the network until it begins to lose its generalization. Then the algorithm measures the sensitivity and automatically prunes away the most irrelevant unit. We define this sensitivity as the absolute difference between the desirable output and the output of the pruned network. Unlike other pruning methods, our algorithm is distinct in calculating the sensitivity from the validation set, instead of the training set, without increasing the asymptotic time complexity of the back-propagation algorithm. In addition, for a classification problem, we raise a point that the sensitivities of some well-known pruning algorithms may still underestimate the irrelevance of hidden unit even though the validation set is used in measuring the sensitivity. We resolve this problem by considering the number of misclassified patterns as the main concern. The Spartan simplicity algorithm is applied to three artificial and seven standard benchmarks. In most problems, the algorithm can produce a compact-sized network with high generalization ability in comparison with other pruning algorithms. © 2008 World Scientific Publishing Company.
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Item,
    Hidden unit reduction of artificial neural network on English capital letter recognition
    (2006-12-01)
    Jearanaitanakij, Kietikul
    ;
    Pinngern, Ouen
    We present an analysis on the minimum number of hidden units that is required to recognize English capital letters of the artificial neural network. The letter font that we use as a case study is the System font. In order to have the minimum number of hidden units, the number of input features has to be minimized. Firstly, we apply our heuristic for pruning unnecessary features from the data set. The small number of the remaining features leads the artificial neural network to have the small number of input units as well. The reason is a particular feature has a one-to-one mapping relationship onto the input unit. Next, the hidden units are pruned away from the network by using the hidden unit pruning heuristic. Both pruning heuristic is based on the notion of the information gain. They can efficiently prune away the unnecessary features and hidden units from the network. The experimental results show the minimum number of hidden units required to train the artificial neural network to recognize English capital letters in System font In addition, the accuracy rate of the classification produced by the artificial neural network is practically high. As a result, the final artificial neural network that we produce is fantastically compact and reliable. © 2006 IEEE.
  • Loading...
    Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Item,
    An information gain technique for acceleration of convergence of artificial neural networks
    (2005-12-01)
    Jearanaitanakij, Kietikul
    ;
    Pinngern, Ouen
    This paper presents an application of information gain to accelerate the convergence time of Artificial Neural Networks (ANNs). We improve Hagiwara's convergence acceleration algorithm by applying information gain to it. The first step of our proposed technique is to calculate information gains of all features (or attributes) in training data and pass those gains through all hidden units in the next layer. During the training process, the algorithm monitors sum-squared error at the output layer. When the variation of sum-squared error becomes small, the worst hidden unit is detected. Next, all the weights connected to the worst hidden unit are reset to random values within the appropriate ranges. These ranges are determined by the propagated information gain of the worst hidden unit. Then, the network is retrained. When the number of weight resetting trials reaches a certain number, a new hidden unit is added to the network and the whole training process is repeated. Our experimental results on standard benchmarks show remarkable outputs in terms of convergence time. © 2005 IEEE.