• Journal of Internet Computing and Services
• /
• v.22 no.1
• /
• pp.23-30
• /
• 2021

This paper proposes a noise-tolerant image classification system using multiple autoencoders. The development of deep learning technology has dramatically improved the performance of image classifiers. However, if the images are contaminated by noise, the performance degrades rapidly. Noise added to the image is inevitably generated in the process of obtaining and transmitting the image. Therefore, in order to use the classifier in a real environment, we have to deal with the noise. On the other hand, the autoencoder is an artificial neural network model that is trained to have similar input and output values. If the input data is similar to the training data, the error between the input data and output data of the autoencoder will be small. However, if the input data is not similar to the training data, the error will be large. The proposed system uses the relationship between the input data and the output data of the autoencoder, and it has two phases to classify the images. In the first phase, the classes with the highest likelihood of classification are selected and subject to the procedure again in the second phase. For the performance analysis of the proposed system, classification accuracy was tested on a Gaussian noise-contaminated MNIST dataset. As a result of the experiment, it was confirmed that the proposed system in the noisy environment has higher accuracy than the CNN-based classification technique.

• Korean Management Science Review
• /
• v.18 no.2
• /
• pp.25-38
• /
• 2001

The well-known back-propagation algorithm for multi-layered neural network has successfully been applied to pattern c1assification problems with remarkable flexibility. Recently. the multi-layered neural network is used as a powerful data mining tool. Nevertheless, in many cases with complex boundary of classification, the successful learning is not guaranteed and the problems of long learning time and local minimum attraction restrict the field application. In this paper, an Improved learning procedure of multi-layered neural network is proposed. The procedure is based on the generalized delta rule but it is particular in the point that the architecture of network is not fixed but enlarged during learning. That is, the number of hidden nodes or hidden layers are increased to help finding the classification boundary and such procedure is controlled by entropy evaluation. The learning speed and the pattern classification performance are analyzed and compared with the back-propagation algorithm.

• Journal of the Korean Recycled Construction Resources Institute
• /
• v.9 no.3
• /
• pp.348-355
• /
• 2021

Asphalt, widely used for road pavement, has different required physical properties depending on the environment to which the road is exposed. Therefore, it is essential to maximize the life of asphalt roads by evaluating the physical properties of asphalt according to additives and selecting an appropriate formulation considering road traffic and climatic environment. Dynamic shear rheometer(DSR) test is mainly used to measure resistance to rutting among various physical properties of asphalt. However, the DSR test has limitations in that the results are different depending on the experimental setting and can only be measured within a specific temperature range. Therefore, in this study, to overcome the limitations of the DSR test, the rheological characteristics were predicted by learning the images collected from atomic force microscopy. Images and rheology properties were trained through EfficientNet, one of the deep learning architectures, and transfer learning was used to overcome the limitation of the deep learning model, which require many data. The trained model predicted the rheological properties of the asphalt binder with high accuracy even though different types of additives were used. In particular, it was possible to train faster than when transfer learning was not used.

• Journal of KIISE:Software and Applications
• /
• v.34 no.12
• /
• pp.1045-1055
• /
• 2007

Latent topic models are statistical models which automatically captures salient patterns or correlation among features underlying a data collection in a probabilistic way. They are gaining an increased popularity as an effective tool in the application of automatic semantic feature extraction from text corpus, multimedia data analysis including image data, and bioinformatics. Among the important issues for the effectiveness in the application of latent topic models to the massive data set is the efficient learning of the model. The paper proposes an accelerated learning technique for PLSA model, one of the popular latent topic models, by an incremental EM algorithm instead of conventional EM algorithm. The incremental EM algorithm can be characterized by the employment of a series of partial E-steps that are performed on the corresponding subsets of the entire data collection, unlike in the conventional EM algorithm where one batch E-step is done for the whole data set. By the replacement of a single batch E-M step with a series of partial E-steps and M-steps, the inference result for the previous data subset can be directly reflected to the next inference process, which can enhance the learning speed for the entire data set. The algorithm is advantageous also in that it is guaranteed to converge to a local maximum solution and can be easily implemented just with slight modification of the existing algorithm based on the conventional EM. We present the basic application of the incremental EM algorithm to the learning of PLSA and empirically evaluate the acceleration performance with several possible data partitioning methods for the practical application. The experimental results on a real-world news data set show that the proposed approach can accomplish a meaningful enhancement of the convergence rate in the learning of latent topic model. Additionally, we present an interesting result which supports a possible synergistic effect of the combination of incremental EM algorithm with parallel computing.

• Journal of the Korean GEO-environmental Society
• /
• v.16 no.5
• /
• pp.43-53
• /
• 2015

In this study, detection methods of outlier in various monitoring data that fit into big data category were developed and outlier detections were conducted for both artificial data and real field monitoring data. Rule-based methods applied rate of change and probability of error for monitoring data are effective to detect a large-scale short faults and constant faults having no change within a certain period. There are however, problems with misjudgement that consider the normal data with a large scale variation as outlier caused by using independent single dataset. Rule-based methods for noise faults detection have a limit to application of real monitoring data due to the problem with a choice of proper window size of data and finding of threshold for outlier judgment. A correlation analysis among different two datasets were very effective to detect localized outlier and abnormal variation for short and long-term monitoring dataset if reasonable range of training data could be selected.

• The Journal of the Korea Contents Association
• /
• v.10 no.3
• /
• pp.472-479
• /
• 2010

The purpose of this paper is to investigates the determinants of learner intention to reuse in e-learning. Based on attitude and flow related studies, our paper proposes a theoretical model consisting of factors such as contents sufficiency, course feedback, self-directed learning, attitude, flow, and reuse intention. The survey method was used for this paper, and data from a total of 409 users in e-learning system were used for this analysis. To analyze the data, structural equation model was used. The results of this empirical study is summarized as follows. First, contents sufficiency has a positive effect on flow as well as attitude, and learner's self-directed learning has a positive effect on flow as well as attitude. Second, both attitude and flow have a positive effect on reuse intention. The findings have significant implications for determinant indicators of reuse intention in e-learning.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.2
• /
• pp.204-211
• /
• 2020

In this paper, we propose an optimal deep learning network architecture for main object location detection through weak supervised learning. The proposed network adds convolution blocks for improving the localization accuracy of the main object through weakly-supervised learning. The additional deep learning network consists of five additional blocks that add a composite product layer based on VGG-16. And the proposed network was trained by the method of weakly-supervised learning that does not require real location information for objects. In addition, Grad-CAM to compensate for the weakness of GAP in CAM, which is one of weak supervised learning methods, was used. The proposed network was tested through the CUB-200-2011 data set, we could obtain 50.13% in top-1 localization error. Also, the proposed network shows higher accuracy in detecting the main object than the existing method.

• KNOM Review
• /
• v.24 no.2
• /
• pp.24-34
• /
• 2021

Service providers using edge computing provide a high level of service. As a result, devices store important information in inner storage and have become a target of the latest cyberattacks, which are more difficult to detect. Although experts use a security system such as intrusion detection systems, the existing intrusion systems have low detection accuracy. Therefore, in this paper, we proposed a machine learning model for more accurate intrusion detections of devices in edge computing. The proposed model is a hybrid model that combines a stacked sparse autoencoder (SSAE) and a convolutional neural network (CNN) to extract important feature vectors from the input data using sparsity constraints. To find the optimal model, we compared and analyzed the performance as adjusting the sparsity coefficient of SSAE. As a result, the model showed the highest accuracy as a 96.9% using the sparsity constraints. Therefore, the model showed the highest performance when model trains only important features.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.23 no.7
• /
• pp.80-86
• /
• 2019

As computer performance improves, research using deep learning are being actively carried out in various fields. Recently, deep learning technology has been applying to the safety evaluation for structures. In particular, the internal blades of a turbine structure requires experienced experts and considerable time to detect surface damages because of the difficulty of separation of the blades from the structure and the dark environmental condition. This study proposes a Faster R-CNN deep learning model that can detect surface damages on the internal blades, which is one of the primary elements of the turbine structure. The deep learning model was trained using image data with dent and punch damages. The image data was also expanded using image filtering and image data generator techniques. As a result, the deep learning model showed 96.1% accuracy, 95.3% recall, and 96% precision. The value of the recall means that the proposed deep learning model could not detect the blade damages for 4.7%. The performance of the proposed damage detection system can be further improved by collecting and extending damage images in various environments, and finally it can be applicable for turbine engine maintenance.

• KIPS Transactions on Software and Data Engineering
• /
• v.5 no.9
• /
• pp.433-440
• /
• 2016

Railway point machines act as actuators that provide different routes to trains by driving switchblades from the current position to the opposite one. Since point failure can significantly affect railway operations with potentially disastrous consequences, early stress detection of point machine is critical for monitoring and managing the condition of rail infrastructure. In this paper, we propose a stress detection method for point machine in railway condition monitoring systems using sound data. The system enables extracting sound feature vector subset from audio data with reduced feature dimensions using feature subset selection, and employs support vector machines (SVMs) for early detection of stress anomalies. Experimental results show that the system enables cost-effective detection of stress using a low-cost microphone, with accuracy exceeding 98%.