• Journal of Intelligence and Information Systems
• /
• v.27 no.3
• /
• pp.139-156
• /
• 2021

The biggest reason for using a deep learning model in image classification is that it is possible to consider the relationship between each region by extracting each region's features from the overall information of the image. However, the CNN model may not be suitable for emotional image data without the image's regional features. To solve the difficulty of classifying emotion images, many researchers each year propose a CNN-based architecture suitable for emotion images. Studies on the relationship between color and human emotion were also conducted, and results were derived that different emotions are induced according to color. In studies using deep learning, there have been studies that apply color information to image subtraction classification. The case where the image's color information is additionally used than the case where the classification model is trained with only the image improves the accuracy of classifying image emotions. This study proposes two ways to increase the accuracy by incorporating the result value after the model classifies an image's emotion. Both methods improve accuracy by modifying the result value based on statistics using the color of the picture. When performing the test by finding the two-color combinations most distributed for all training data, the two-color combinations most distributed for each test data image were found. The result values were corrected according to the color combination distribution. This method weights the result value obtained after the model classifies an image's emotion by creating an expression based on the log function and the exponential function. Emotion6, classified into six emotions, and Artphoto classified into eight categories were used for the image data. Densenet169, Mnasnet, Resnet101, Resnet152, and Vgg19 architectures were used for the CNN model, and the performance evaluation was compared before and after applying the two-stage learning to the CNN model. Inspired by color psychology, which deals with the relationship between colors and emotions, when creating a model that classifies an image's sentiment, we studied how to improve accuracy by modifying the result values based on color. Sixteen colors were used: red, orange, yellow, green, blue, indigo, purple, turquoise, pink, magenta, brown, gray, silver, gold, white, and black. It has meaning. Using Scikit-learn's Clustering, the seven colors that are primarily distributed in the image are checked. Then, the RGB coordinate values of the colors from the image are compared with the RGB coordinate values of the 16 colors presented in the above data. That is, it was converted to the closest color. Suppose three or more color combinations are selected. In that case, too many color combinations occur, resulting in a problem in which the distribution is scattered, so a situation fewer influences the result value. Therefore, to solve this problem, two-color combinations were found and weighted to the model. Before training, the most distributed color combinations were found for all training data images. The distribution of color combinations for each class was stored in a Python dictionary format to be used during testing. During the test, the two-color combinations that are most distributed for each test data image are found. After that, we checked how the color combinations were distributed in the training data and corrected the result. We devised several equations to weight the result value from the model based on the extracted color as described above. The data set was randomly divided by 80:20, and the model was verified using 20% of the data as a test set. After splitting the remaining 80% of the data into five divisions to perform 5-fold cross-validation, the model was trained five times using different verification datasets. Finally, the performance was checked using the test dataset that was previously separated. Adam was used as the activation function, and the learning rate was set to 0.01. The training was performed as much as 20 epochs, and if the validation loss value did not decrease during five epochs of learning, the experiment was stopped. Early tapping was set to load the model with the best validation loss value. The classification accuracy was better when the extracted information using color properties was used together than the case using only the CNN architecture.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2014.10a
• /
• pp.342-345
• /
• 2014

Network intrusion detection system based on machine learning methods is quite dependent on the selected features in terms of accuracy and efficiency. Nevertheless, choosing the optimal combination of features from generally used features to detect network intrusion requires extensive computing resources. For instance, the number of possible feature combinations from given n features is $2^n-1$. In this paper, to tackle this problem we propose a optimal feature selection algorithm. Proposed algorithm is based on the local search algorithm, one of representative meta-heuristic algorithm for solving optimization problem. In addition, the accuracy of clusters which obtained using selected feature components and k-means clustering algorithm is adopted to evaluate a feature assembly. In order to estimate the performance of our proposed algorithm, comparing with a method where all features are used on NSL-KDD data set and multi-layer perceptron.

• The KIPS Transactions:PartB
• /
• v.15B no.4
• /
• pp.365-374
• /
• 2008

In this paper, we propose a model to enrich an ontology by incrementally extending the relations through variations of patterns. In order to generalize initial patterns, combinations of features are considered as candidate patterns. The candidate patterns are used to extract relations from Wikipedia, which are sorted out according to reliability based on corpus frequency. Selected patterns then are used to extract relations, while extracted relations are again used to extend the patterns of the relation. Through making variations of patterns in incremental enrichment process, the range of pattern selection is broaden and refined, which can increase coverage and accuracy of relations extracted. In the experiments with single-feature based pattern models, we observe that the features of lexical, headword, and hypernym provide reliable information, while POS and syntactic features provide general information that is useful for enrichment of relations. Based on observations on the feature types that are appropriate for each syntactic unit type, we propose a pattern model based on the composition of features as our ongoing work.

• Journal of the Korean BIBLIA Society for library and Information Science
• /
• v.22 no.2
• /
• pp.5-25
• /
• 2011

This study aims to identify the most effective statistical feature selecting method and context window size for word sense disambiguation using supervised methods. In this study, features were selected by four different methods: information gain, document frequency, chi-square, and relevancy. The result of weight comparison showed that identifying the most appropriate features could improve word sense disambiguation performance. Information gain was the highest. SVM classifier was not affected by feature selection and showed better performance in a larger feature set and context size. Naive Bayes classifier was the best performance on 10 percent of feature set size. kNN classifier on under 10 percent of feature set size. When feature selection methods are applied to word sense disambiguation, combinations of a small set of features and larger context window size, or a large set of features and small context windows size can make best performance improvements.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.17 no.2
• /
• pp.642-648
• /
• 2016

As 3D solid modeling prevails, a range of applications have become possible and intensive research on the integration of CAD/CAM has been conducted. As a consequence, methods to recognize the machining features from CAD models have been developed. On the other hand, generating a machining sequence using the machining features is still a problem due to a combinatorial problem with a large number of machining features. This paper proposes a new method that utilizes the precedence constraints through which the number of the combinations is reduced drastically. This method can automatically generate machining sequences requiring the lowest amount of machining time. An airplane part was used to test the usefulness of the proposed method.

• Korean Journal of Clinical Laboratory Science
• /
• v.42 no.2
• /
• pp.97-102
• /
• 2010

The two new female cases of Angelman syndrome (AS) were described, which diagnosed on the basis of clinical features (dysmorphic facial features, severe mental retardation with absent speech, peculiar jerky movements, ataxic gait and paroxysms of inappropriate laughter) and neurophysiological findings. Failure to detect the deletion of the long arm of chromosome 15 or the absence of epileptic seizure were not considered sufficient to exclude a diagnosis of AS. Feeding problems, developmental delay and early signs of ataxia, especially tremor on handling objects and unstable posture when seated, proved effective as the clinical markers for early diagnosis of AS. Most of the authors agreed about the existence of three main EEG patterns in AS which may appear in isolation or in various combinations in the same patient. The most frequently observed pattern in children has prolonged runs of high amplitude rhythmic 2-3 Hz activity predominantly over the frontal region with superimposed interictal epileptiform discharges. High amplitude rhythmic 4-6 Hz activity, prominent in the occipital regions, with spikes, which can be facilitated by eye closure, is often seen in children under the age of 12 years. The EEG findings are characteristic of AS when seen in the appropriate clinical context and can be helpful to identify AS patients at an early age when genetic counselling may be particularly important.

• Proceedings of the KSRS Conference
• /
• v.1
• /
• pp.138-141
• /
• 2006

In developing algorithms to retrieve the sea surface temperature (SST) and sea surface wind speed from the Advanced Microwave Scanning Radiometer (AMSR) aboard the AQUA and the Advanced Earth Observation Satellite-II (ADEOS-II), data from the SeaWinds aboard ADEOS-II were helpful. Since features of the ocean microwave emission (Tb) related with ocean wind are not well understood, in case of using only AMSR data, combination of AMSR and SeaWinds revealed pronounced features about the ocean Tb. Two results from combinations of the two sensors were shown in this paper. One result was obtained at wind speeds over about 6m/s, in which the ocean Tb varies with the air-sea temperature difference, even though the SeaWinds wind speed is fixed at the same values. The ocean Tb increases as the air-sea temperature difference becomes negative, i.e., the boundary condition becomes unstable. This result indicates that the air temperature should be included in AMSR SST algorithm. The second result was obtained from comparison of two wind speeds between AMSR and SeaWinds. There is a small difference of two wind speeds, which might be related with several mechanisms, such as evaporation and plankton.

• Journal of the Korean housing association
• /
• v.21 no.3
• /
• pp.67-76
• /
• 2010

The purposes of this study are to understand residents' needs in regard to living space and to suggest how to provide layout options for the infill, based on their needs, so that the residents can change their living space to suit their own need. This study analyzed residents' needs in terms of living spaces through literature reviews on apartment remodeling and related previous studies. The results are as follows: First, the residents remodeled the various infill, and remodeling works are then classified into five infill groups according to the flexible features: 1) structural elements, such as flooring, ceilings, interior walls, and windows/doors; 2) equipment elements, such as lighting and electricity, electrical wiring, heating arrangements, and water supply & drainage systems: 3) finishing material elements, such as finishing materials for floors, walls, and ceilings, skirting boards, moldings, and art walls; 4) furniture elements, such as built-in wardrobes, storage closets, and kitchen cabinets; and 5) bathroom facility elements such as faucets and sinks. Second, based on the remodeling features, four ways to provide options can be suggested. 1) options are provided for each room; 2) options are provided in connection with structural elements; 3) options are provided for each finishing material element; and 4) options are provided with the combinations of different bathroom facilities.

• Journal of Korea Multimedia Society
• /
• v.9 no.3
• /
• pp.287-295
• /
• 2006

This paper compares two different approaches based on wavelet and Gabor decomposition towards representing the texture of an arbitrary region. The Gabor-domain mean and standard deviation combination is considered to be best in representing the texture of rectangular regions. However, texture representation of arbitrary regions would enable generalized object-based image retrieval and other applications in the future. In this study, we have found that the wavelet features perform better than the Gabor features in representing the texture of arbitrary regions. Particularly, the wavelet-domain standard deviation and entropy combination results in the best retrieval accuracy. Based on our experiment with texture image sets, we present and compare tile retrieval accuracy of multiple wavelet and Gabor feature combinations.

• Journal of the Korea Safety Management & Science
• /
• v.12 no.4
• /
• pp.207-213
• /
• 2010

Now days, the marketing environment is rapidly changed. Therefore, a lot of companies try to reduce production cost. Especially, Design is a important activities in new product development. While the concepts of design for manufacturable and concurrent engineering have made significant advances in integrating the design function with other areas in the firm. There are still major gaps in timely and accurate costing information available to designers. Inappropriate design could result in high redesign cost and delay in product relation. The generation of design and improvement is a time-consuming and mentally exhaustion process. It involves combining design features to generate as many potential design as possible. As not all features combinations are feasible, decision-makers have to narrow down the potential solutions and subsequently select appropriate design for further development. This new study is composed of 3 steps aiming at the Low Cost Design of the product. The three steps are consisted that setting up the target Cost, estimating the current functional cost, the design of a unit's reviewing according to the priority of the difference between the target cost and the functional cost.