• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.27 no.1
• /
• pp.8-14
• /
• 2023

The existing self-supervised learning-based CutPaste generated synthetic data by cutting and attaching specific patches from normal images and then performed anomaly detection. However, this method has a problem in that there is a clear difference in the boundary of the patch. NSA for solving these problems have achieved higher anomaly detection performance by generating natural synthetic data through Poisson Blending. However, NSA has the disadvantage of having many hyperparameters that need to be adjusted for each class. In this paper, synthetic data similar to normal were generated by a simple method of making the size of the synthetic patch very small. At this time, since the patches are so locally synthesized, models that learn local features can easily overfit synthetic data. Therefore, we performed anomaly detection using gMLP, which learns global features, and even with simple synthesis methods, we were able to achieve higher performance than conventional self-supervised learning techniques.

• The Journal of the Acoustical Society of Korea
• /
• v.23 no.1
• /
• pp.1-7
• /
• 2004

We developed SH-SAW sensors to detect protein molecules in liquid solutions applying a particular antibody thin film on the delay line of transverse SAW devices. The antibody investigated was human-immune-globulin G (HigG) to hold the antigens (anti-HigG) in the protein solution. We fabricated the sensor generating 100 MHz with the piezoelectric single crystal LiTaO₃. We measured the frequency change of the sensor by adding the anti-body concentration on SAM (self assembled monolayer) deposited on the Au layer. The sensor showed stable response to the mass loading effects of the anti-HigG molecules with the sensitivity up to 10.8 ng/ml/Hz at noise level 400 Hz below.

• Journal of Korean Institute of Industrial Engineers
• /
• v.19 no.3
• /
• pp.123-137
• /
• 1993

In this paper, an integrated approach is proposed to generate gouging-free Cartesian tool paths for machining sculptured surfaces from 3D measurement data. The integrated CAD/CAM system consists of two modules : offset surface module an Carteian tool path module. The offset surface module generates an offset surface of an object from its 3D measurement data, using an offsetting method and a surface fitting method. The offsetting is based on the idea that the envelope of an inversed tool generates an offset surface without self-intersection as the center of the inversed tool moves along on the surface of an object. The surface-fitting is the process of constructing a compact representation to model the surface of an object based on a fairly large number of data points. The resulting offset surtace is a composite Bezier surface without self-intersection. When an appropriate tool-approach direction is selected, the tool path module generates the Cartesian tool paths while the deviation of the tool paths from the surface stays within the user-specified tolerance. The tool path module is a two-step process. The first step adaptively subdivides the offset surface into subpatches until the thickness of each subpatch is small enough to satisfy the user-defined tolerance. The second step generates the Cartesian tool paths by calculating the intersection of the slicing planes and the adaptively subdivided subpatches. This tool path generation approach generates the gouging-free Cartesian CL tool paths, and optimizes the cutter movements by minimizing the number of interpolated points.

• Journal of Korean Elementary Science Education
• /
• v.30 no.3
• /
• pp.305-314
• /
• 2011

In this study, the analogies on saturated solution generated by elementary school teachers were analyzed in their numbers, materials, and types aspects. The results were also compared with those of general and science-gifted elementary students. A test on the self-generating analogies on the target concept was administered to 111 elementary school teachers, 60 fifth graders at four science-gifted education institutes and 91 fifth graders at three elementary schools. The results revealed that the teachers made more analogies than the general and science-gifted students. In general, both the teachers and the students tended to make the analogies using the materials in family, riding, digestive, and school situations. However, there were a little difference between the teachers and the students in the cases of the analogies using other situations including body/physical activity. Similarly to the cases of the students, the teachers made more functional analogies than structural or structural/functional ones and did more concrete ones than abstract ones. However, they made more verbal, artificial, and enriched analogies than the students. They also made more highly systematic analogies than the general students, and did less ones than the science-gifted students. Educational implications of these findings are discussed.

• Bulletin of the Korean Chemical Society
• /
• v.32 no.10
• /
• pp.3730-3734
• /
• 2011

The degree of self-assembly and the size variation of nanotubular structures in anodic titanium oxide prepared by the anodization of titanium in ethylene glycol containing 0.25 wt % $NH_4F$ at 40 V were investigated as a function of anodization time. We found that the degree of self-assembly and the size of the nanotubes were strongly dependent on thickness deviation and thus indirectly on anodization time, as the thickness deviation was caused by the dissolution of the topmost tubular structures at local areas during long anodization. A large deviation in thickness led to a large deviation in the size and number of nanotubes per unit area. The dissolution primarily occurred at the bottoms of the nanotubes ($D_{bottom}$) in the initial stage of anodization (up to 6 h), which led to the growth of nanotubes. Dissolution at the tops ($D_{top}$) was accompanied by $D_{bottom}$ after the formed structures contacted the electrolyte after 12 h, generating the thickness deviation. After extremely long anodization (here, 70 h), $D_{top}$ was the dominant mode due to increase in pH, meaning that there was insufficient driving force to overcome the size distribution of nanotubes at the bottom. Thus, the nanotube array became disorder in this regime.

• Korean Institute of Interior Design Journal
• /
• v.26 no.1
• /
• pp.13-24
• /
• 2017

The semiotics philosopher Roland Barthes declared 'the death of the author'. This means a sense of skepticism about author-centered writing in literature, and that readers and critics must also change their reading attitudes. In architecture, too, author-centered design appears. Architects sometimes try to express their will and concepts strongly through their work. However, this architecture is oriented toward identity, self-revealing, self-centered, and eventually without inspiration. On the other hand, the signifier does not result in a single meaning, and it is architecture as text that allows readers to enjoy the pleasure of deciphering. The purpose of this study is to investigate that unilateral by the will of the architect. Or whether there is a potential for non-intelligence and the possibility of generating meaning. In particular, we will analyze the works of Rem Koolhaas, SANAA and Sou Fujimoto to see if they are free from the will of the architect and can obtain the user's true freedom. As a result, we want to reveal the potential of various interpretations by users in contemporary architecture.

• International Journal of Contents
• /
• v.17 no.4
• /
• pp.1-15
• /
• 2021

Named entity recognition (NER) is a crucial task for NLP, which aims to extract information from texts. To build NER systems, deep learning (DL) models are learned with dictionary features by mapping each word in the dataset to dictionary features and generating a unique index. However, this technique might generate noisy labels, which pose significant challenges for the NER task. In this paper, we proposed DL-dictionary features, and evaluated them on two datasets, including the OntoNotes 5.0 dataset and our new infectious disease outbreak dataset named GFID. We used (1) a Bidirectional Long Short-Term Memory (BiLSTM) character and (2) pre-trained embedding to concatenate with (3) our proposed features, named the Convolutional Neural Network (CNN), BiLSTM, and self-attention dictionaries, respectively. The combined features (1-3) were fed through BiLSTM - Conditional Random Field (CRF) to predict named entity classes as outputs. We compared these outputs with other predictions of the BiLSTM character, pre-trained embedding, and dictionary features from previous research, which used the exact matching and partial matching dictionary technique. The findings showed that the model employing our dictionary features outperformed other models that used existing dictionary features. We also computed the F1 score with the GFID dataset to apply this technique to extract medical or healthcare information.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.19-21
• /
• 2022

Anomaly detection using self-supervised learning typically generates synthetic data to learn to classify normal and abnormal, and uses real abnormal data as test data to measure anomaly detection performance. In a study using this method to generate synthetic data similar to normal data, anomaly detection was carried out by generating synthetic data by cutting and pasting a specific patch from the original image. In this way, the degree of similarity to normal data depends on the number and size of patches, which affects anomaly detection performance. In this paper, synthetic data were generated by varying patch sizes and numbers, and then similarity and analysis with normal data were conducted using a pre-trained model, and anomaly detection performance was measured by learning the model.

• IMMUNE NETWORK
• /
• v.19 no.1
• /
• pp.6.1-6.14
• /
• 2019

Plasmacytoid dendritic cells (pDCs) are a unique subset of cells with different functional characteristics compared to classical dendritic cells. The pDCs are critical for the production of type I IFN in response to microbial and self-nucleic acids. They have an important role for host defense against viral pathogen infections. In addition, pDCs have been well studied as a critical player for breaking tolerance to self-nucleic acids that induce autoimmune disorders such as systemic lupus erythematosus. However, pDCs have an immunoregulatory role in inducing the immune tolerance by generating Tregs and various regulatory mechanisms in mucosal tissues. Here, we summarize the recent studies of pDCs that focused on the functional characteristics of gut pDCs, including interactions with other immune cells in the gut. Furthermore, the dynamic role of gut pDCs will be investigated with respect to disease status including gut infection, inflammatory bowel disease, and cancers.

• Journal of The Korean Association For Science Education
• /
• v.37 no.3
• /
• pp.407-416
• /
• 2017

In this study, we investigated the processes of analogy change and the characteristics of discussions in collaborative activity of self-generated analogy. Twenty-four high school students in Seoul participated in this study. We implemented science instructions based on collaborative activity of self-generated analogy. We compared personal analogies, group analogies, and modified group analogies in order to analyze the processes of analogy change. We also analyzed the characteristics of group and classroom discussions in the science instructions. The analyses of the results indicated that the processes of analogy change were categorized into three patterns; adding shared attributes, recognizing unshared attributes, and revising mapping errors. They selected a group analogy from analogies of their group members by considering inclusiveness, originality, and familiarity. They perceived the activity of self-generated analogy as subjective and creative. Therefore, they felt little pressure of self-generated analogy and there were little conflicts in group discussions. On the other hand, various analogies were suggested in classroom discussions and the competitive atmosphere of classroom discussions led students to focus on unshared attributes. At the stage of modifying group analogies, they added unshared attributes as limitations of the group analogy and changed their group analogy not to have unshared attributes. There were no cases of generating a new analogy. Some suggestions to implement collaborative activity of self-generated analogy in science teaching effectively are discussed.