• The Magazine of Kiice
• /
• v.18 no.1
• /
• pp.43-47
• /
• 2017

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2009.06a
• /
• pp.945-946
• /
• 2009

• Korean Journal of Artificial Intelligence
• /
• v.5 no.1
• /
• pp.29-37
• /
• 2017

The range of problems that can be handled by the activation of big data and the development of hardware has been rapidly expanded and machine learning such as deep learning has become a very versatile technology. In this paper, mnist data set is used as experimental data, and the Cross Entropy function is used as a loss model for evaluating the efficiency of machine learning, and the value of the loss function in the steepest descent method is We applied the Gradient Descent Optimize algorithm to minimize and updated weight and bias via backpropagation. In this way we analyze optimal reliability value corresponding to the number of exercises and optimal reliability value without overfitting. And comparing the overfitting time according to the number of data changes based on the number of training times, when the training frequency was 1110 times, we obtained the result of 92%, which is the optimal reliability value without overfitting.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.9 no.5
• /
• pp.472-479
• /
• 1999

In this paper, we propose a land cover pattern classifier for remote sensing image by using neuro-fuzzy algorithm. The proposed pattem classifier has a 3-layer feed-forward architecture that is derived from generic fuzzy perceptrons, and the weights are con~posed of h u y sets. We also implement a neuro-fuzzy pattern classification system in the Visual C++ environment. To measure the performance of this, we compare it with the conventional neural networks with back-propagation learning and the Maximum-likelihood algorithms. We classified the remote sensing image into the eight classes covered the majority of land cover feature, selected the same training sites. Experimental results show that the proposed classifier performs well especially in the mixed composition area having many classes rather than the conventional systems.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.2
• /
• pp.343-350
• /
• 2022

It is worth verifying the effectiveness of data integration between data with different features. This study investigated whether the data integration affects the accuracy of deep neural network (DNN), and which integration method shows the best improvement. This study used two different public datasets. One public dataset was taken in an actual farm in India. And another was taken in a laboratory environment in Korea. Leaf images were selected from two different public datasets to have five classes which includes normal and four different types of plant diseases. DNN used pre-trained VGG16 as a feature extractor and multi-layer perceptron as a classifier. Data were integrated into three different ways to be used for the training process. DNN was trained in a supervised manner via the integrated data. The trained DNN was evaluated by using a test dataset taken in an actual farm. DNN shows the best accuracy for the test dataset when DNN was first trained by images taken in the laboratory environment and then trained by images taken in the actual farm. The results show that data integration between plant images taken in a different environment helps improve the performance of deep neural networks. And the results also confirmed that independent use of plant images taken in different environments during the training process is more effective in improving the performance of DNN.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.13 no.6
• /
• pp.635-642
• /
• 2003

Reinforcement learning got successful results in a lot of applications such as control and scheduling. Various function approximation methods have been studied in order to improve the learning speed and to solve the shortage of storage in the standard reinforcement learning algorithm of Q-Learning. Most function approximation methods remove some special quality of reinforcement learning and need prior knowledge and preprocessing. Fuzzy Q-Learning needs preprocessing to define fuzzy variables and Local Weighted Regression uses training examples. In this paper, we propose a function approximation method, Fuzzy Q-Map that is based on on-line fuzzy clustering. Fuzzy Q-Map classifies a query state and predicts a suitable action according to the membership degree. We applied the Fuzzy Q-Map, CMAC and LWR to the mountain car problem. Fuzzy Q-Map reached the optimal prediction rate faster than CMAC and the lower prediction rate was seen than LWR that uses training example.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.5
• /
• pp.631-638
• /
• 2012

This work proposes a new method to generate velocity profile of a traction motor equipped in a rehabilitation system for knee joint patients through humanoid robot simulation. To this end, a three-dimensional full-body humanoid robot model is newly constructed, and natural human gait is simulated by applying to it reference joint angle trajectories already published. Linear velocity is derived from distance data calculated between the positions of a thigh band and its traction motor at every sampling instance, which is a novel idea of this paper. The projection rule is employed to kinematically describe the humanoid robot because of its high efficiency and accuracy, and measured joint trajectories are used in simulating human natural gait referring to Winter's book. The attained motor velocity profile for a certain position in human body will be applied to our walking-assistance system which is implemented with a treadmill system.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.18 no.6
• /
• pp.768-774
• /
• 2008

This paper implements a information management system for a bullfights using RFID technology in which the objects containing electronic tag are automatically identified by using radio frequency wave. The presented system is composed of the information systems for career and train, and the real-time informant system to provide various informations to remote bull breeders and administers. As the first step for implementation, we analyze the requirements for information management systems based on the RFID and suggests design considerations. Based on the analysis, we implement an efficient RFID middleware system for tracking a tag location based on antenna transfer method and managing an intelligent tag information which efficiently manages a bullfight informations in field. And we also implement the web-based integrated management system for managing and providing a bullfight informations. The career management system sequentially recognizes a tags, and the training system concurrently recognizes the man tags.

• The Journal of the Korea Contents Association
• /
• v.22 no.9
• /
• pp.46-63
• /
• 2022

With the rapid development of digital technology and the growing trend to integrate it into the medical field, recent studies suggest language rehabilitation for people with aphasia using virtual reality (VR), augmented reality (AR) and mobile applications. This study conducted a scoping review to summarize the features of digital technology-based language rehabilitation for aphasia in the last four years (2018-2021) and draw implications for future research. A total of 20 papers met the selection criteria among the documents retrieved from the Web of Science, CINAHL, and RISS. This review demonstrates that digital technology could offer unique treatment content by gamification, individualization, and creating a realistic communication environment, and by utilizing them in various ways. Therefore, we expect digital technology-based language rehabilitation for aphasia could supplement the limitations of conventional language rehabilitation and provide a novel perspective on development of treatment content.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.9
• /
• pp.191-203
• /
• 2022

In this paper, we developed a system that intelligently identifies skin image data from big data collected from social media Instagram and extracts standardized skin sample data for skin condition diagnosis and management. The system proposed in this paper consists of big data collection and analysis stage, skin image analysis stage, training data preparation stage, artificial neural network training stage, and skin image identification stage. In the big data collection and analysis stage, big data is collected from Instagram and image information for skin condition diagnosis and management is stored as an analysis result. In the skin image analysis stage, the evaluation and analysis results of the skin image are obtained using a traditional image processing technique. In the training data preparation stage, the training data were prepared by extracting the skin sample data from the skin image analysis result. And in the artificial neural network training stage, an artificial neural network AnnSampleSkin that intelligently predicts the skin image type using this training data was built up, and the model was completed through training. In the skin image identification step, skin samples are extracted from images collected from social media, and the image type prediction results of the trained artificial neural network AnnSampleSkin are integrated to intelligently identify the final skin image type. The skin image identification method proposed in this paper shows explain high skin image identification accuracy of about 92% or more, and can provide standardized skin sample image big data. The extracted skin sample set is expected to be used as standardized skin image data that is very efficient and useful for diagnosing and managing skin conditions.