• The Korean Journal of Applied Statistics
• /
• v.37 no.5
• /
• pp.631-641
• /
• 2024

Multi-task learning is a popular approach in machine learning that aims to learn multiple related tasks simultaneously by sharing information across them. In this paper, we consider a hierarchical structure across multiple related tasks with a hierarchy of sub-tasks under the same main task, where representations used to solve the sub-tasks share more information through task-specific layers, globally shared layers, and locally shared layers. We thus propose the hierarchical multi-task learning with self-supervised auxiliary task (HiSS), which is a novel approach for hierarchical multi-task learning that incorporates self-supervised learning as an auxiliary task. The goal of the auxiliary task is to further extract latent information from the unlabeled data by predicting a cluster label directly derived from the data. The proposed approach is tested on the Hyodoll dataset, which consists of user information and activity logs of elderly individuals collected by AI companion robots, for predicting emergency calls based on the time of day and month. Our proposed algorithm is more efficient than other well-known machine learning algorithms as it requires only a single model regardless of the number of tasks, and demonstrates superior performance in classification tasks using various metrics. The source codes are available at: https://github.com/seunghan96/HiSS.

• Journal of The Korean Association For Science Education
• /
• v.44 no.5
• /
• pp.361-378
• /
• 2024

This study explores elementary teachers' PCK change through the knowledge exchange between ePCK (enacted PCK) and pPCK (personal PCK) during modeling-based science inquiry instruction. For this purpose, three elementary school teachers were selected as research participants, and the CoRes(content representations) written by the participants before and after the lessons. The CoRes were, then, analyzed using qualitative content analysis to explore changes in pPCK, and the data related to the modeling-based science inquiry instruction (such as lesson plan, lesson videos, in-depth interviews) were analyzed using grounded theory research methods. The research indicated that the pPCK of elementary teachers who enacted modeling-based science inquiry instruction changed in regard to curricular knowledge, knowledge of students, pedagogical knowledge, and assessment knowledge. The data from the modeling-based science inquiry instruction showed that the changes of the teachers' pPCK was influenced by the knowledge exchange between ePCK and pPCK during the process of planning, teaching, and reflecting on the lessons. Based on the findings, this study discussed the need to identify personal factors that influence elementary teachers' knowledge exchange between PCK realms and suggested the need to develop effective modeling-based science inquiry teaching methods.

• Journal of the Daesoon Academy of Sciences
• /
• v.49
• /
• pp.157-186
• /
• 2024

This study aims to understand the symbolic meanings of dragons in the contemporary period through the painting featuring poetry based on the Hwaryong (paintings of dragons) and to analyze the literary representations and characteristics of dragons. First, the shape and aura of the dragon in the paintings should be confirmed. In dragon Painting Poetry (Paintings Featuring Poetry), the shape of the dragon is embodied in the aura of 'spirit.' However, people in previous eras customarily believed that dragons were real. In addition, the main spirit of the Hwaryong was in the representation of the dragon's aura, or "Jeonsin (transmission of spirit)," such as the creature's dragonly virtue or Heavenly Perfection. Next, the rhetorical representation of the dragon in the painting is to be examined. In the actual Painting Poetry, dragon images were relatively common in descriptions of pines, bamboo, and plum trees. The pine-bamboo-plum motif is known as Sehan Samwu (the three friends of the cold season), and since the Northern Song Dynasty, it has become a symbol of literary incision. Manifested as a dragon, the pine-bamboo-plum motif has a common denominator of winding branches, tangled stems, and bluish jade coloration, and like the dragon, it is used as a backdrop for weather phenomena such as rain, wind, and lightning. In the schematic that connects the Hwaryong, dragon Painting Poetry, the original meaning of the dragon is transformed into shapes and language. This is the meaning conveyed by the dragon, and this symbolic meaning was shared among the writers of the period who viewed the poems and paintings.

• Journal of the Korean Society of Radiology
• /
• v.18 no.5
• /
• pp.557-566
• /
• 2024

Ultrasound-guided regional anesthesia is one of the most common techniques used in peripheral nerve blockade by enhancing pain control and recovery time. However, accurate Brachial Plexus (BP) nerve detection and identification remains a challenging task due to the difficulty in data acquisition such as speckle and Doppler artifacts even for experienced anesthesiologists. To mitigate the issue, we introduce a BP nerve small target segmentation network by incorporating BP object detection and U-Net based semantic segmentation into a single deep learning framework based on the multi-scale approach. To this end, the current BP detection and identification was estimated: 1) A RetinaNet model was used to roughly locate the BP nerve region using multi-scale based feature representations, and 2) U-Net was then used by feeding plural BP nerve features for each scale. The experimental results demonstrate that our proposed model produces high quality BP segmentation by increasing the accuracies of the BP nerve identification with the assistance of roughly locating the BP nerve area compared to competing methods such as segmentation-only models.

• Journal of Intelligence and Information Systems
• /
• v.24 no.1
• /
• pp.205-225
• /
• 2018

Convolutional Neural Network (ConvNet) is one class of the powerful Deep Neural Network that can analyze and learn hierarchies of visual features. Originally, first neural network (Neocognitron) was introduced in the 80s. At that time, the neural network was not broadly used in both industry and academic field by cause of large-scale dataset shortage and low computational power. However, after a few decades later in 2012, Krizhevsky made a breakthrough on ILSVRC-12 visual recognition competition using Convolutional Neural Network. That breakthrough revived people interest in the neural network. The success of Convolutional Neural Network is achieved with two main factors. First of them is the emergence of advanced hardware (GPUs) for sufficient parallel computation. Second is the availability of large-scale datasets such as ImageNet (ILSVRC) dataset for training. Unfortunately, many new domains are bottlenecked by these factors. For most domains, it is difficult and requires lots of effort to gather large-scale dataset to train a ConvNet. Moreover, even if we have a large-scale dataset, training ConvNet from scratch is required expensive resource and time-consuming. These two obstacles can be solved by using transfer learning. Transfer learning is a method for transferring the knowledge from a source domain to new domain. There are two major Transfer learning cases. First one is ConvNet as fixed feature extractor, and the second one is Fine-tune the ConvNet on a new dataset. In the first case, using pre-trained ConvNet (such as on ImageNet) to compute feed-forward activations of the image into the ConvNet and extract activation features from specific layers. In the second case, replacing and retraining the ConvNet classifier on the new dataset, then fine-tune the weights of the pre-trained network with the backpropagation. In this paper, we focus on using multiple ConvNet layers as a fixed feature extractor only. However, applying features with high dimensional complexity that is directly extracted from multiple ConvNet layers is still a challenging problem. We observe that features extracted from multiple ConvNet layers address the different characteristics of the image which means better representation could be obtained by finding the optimal combination of multiple ConvNet layers. Based on that observation, we propose to employ multiple ConvNet layer representations for transfer learning instead of a single ConvNet layer representation. Overall, our primary pipeline has three steps. Firstly, images from target task are given as input to ConvNet, then that image will be feed-forwarded into pre-trained AlexNet, and the activation features from three fully connected convolutional layers are extracted. Secondly, activation features of three ConvNet layers are concatenated to obtain multiple ConvNet layers representation because it will gain more information about an image. When three fully connected layer features concatenated, the occurring image representation would have 9192 (4096+4096+1000) dimension features. However, features extracted from multiple ConvNet layers are redundant and noisy since they are extracted from the same ConvNet. Thus, a third step, we will use Principal Component Analysis (PCA) to select salient features before the training phase. When salient features are obtained, the classifier can classify image more accurately, and the performance of transfer learning can be improved. To evaluate proposed method, experiments are conducted in three standard datasets (Caltech-256, VOC07, and SUN397) to compare multiple ConvNet layer representations against single ConvNet layer representation by using PCA for feature selection and dimension reduction. Our experiments demonstrated the importance of feature selection for multiple ConvNet layer representation. Moreover, our proposed approach achieved 75.6% accuracy compared to 73.9% accuracy achieved by FC7 layer on the Caltech-256 dataset, 73.1% accuracy compared to 69.2% accuracy achieved by FC8 layer on the VOC07 dataset, 52.2% accuracy compared to 48.7% accuracy achieved by FC7 layer on the SUN397 dataset. We also showed that our proposed approach achieved superior performance, 2.8%, 2.1% and 3.1% accuracy improvement on Caltech-256, VOC07, and SUN397 dataset respectively compare to existing work.

• The KIPS Transactions:PartA
• /
• v.11A no.7 s.91
• /
• pp.555-562
• /
• 2004

Both global volume reduction and local shape changes of hippocampus within the brain indicate their abnormal neurological states. Hippocampal shape analysis consists of two main steps. First, construct a hippocampal shape representation model ; second, compute a shape similarity from this representation. This paper proposes a novel method for the analysis of hippocampal shape using integrated Octree-based representation, containing meshes, voxels, and skeletons. First of all, we create multi-level meshes by applying the Marching Cube algorithm to the hippocampal region segmented from MR images. This model is converted to intermediate binary voxel representation. And we extract the 3D skeleton from these voxels using the slice-based skeletonization method. Then, in order to acquire multiresolutional shape representation, we store hierarchically the meshes, voxels, skeletons comprised in nodes of the Octree, and we extract the sample meshes using the ray-tracing based mesh sampling technique. Finally, as a similarity measure between the shapes, we compute $L_2$ Norm and Hausdorff distance for each sam-pled mesh pair by shooting the rays fired from the extracted skeleton. As we use a mouse picking interface for analyzing a local shape inter-actively, we provide an interaction and multiresolution based analysis for the local shape changes. In this paper, our experiment shows that our approach is robust to the rotation and the scale, especially effective to discriminate the changes between local shapes of hippocampus and more-over to increase the speed of analysis without degrading accuracy by using a hierarchical level-of-detail approach.

• Cartoon and Animation Studies
• /
• s.38
• /
• pp.53-70
• /
• 2015

In industrialized societies throughout the 19th and 20th centuries, Over Produced and Mass consumption images were constantly shown to people via Mass-Media as means to provoke one's desire. Frank Mouris, the American independent animator, captured and showed the infinite nesting of industrialized image with his autobiographical story through his work (1973) and made it as an intense visual flow. This innovative art animation has broke the traditional form of narrative animation and won the Annecy Animation Festival Grand Prix and the Academy Awards in 1974. This was also selected for preservation in the United States National Film Registry by the Library of Congress as being culturally, historically, or aesthetically significant in 1996. This study explores and shows that how these a-half million images to express Franks Mouris's autobiographical story in could be analyzed by the concept of Walter Benjamin's 'dialectical images'. Typically, the term 'dialectic' need to be formed by contradiction or opposite concept in the basic principles, but a dialectical image of Benjamin could be formed without any opposite concept while maintaining the uniqueness of each new relationship of the past. Benjamin's dialectical images are no longer stay in the historical past, It always meets with the present when someone realizes the past in the present moment. I suggest three different aspect according to Benjamin's point of view to analyse this animated film such as 'Historical-dialectical imaging of private/collective memory', 'Reconfiguring of present through analysing the relationship between the image flows and its own time/space', and 'Old future over the existing fragment and the presence of fragment. has the great value not only to present the experimental and innovative aesthetics of animated film, but also to show an analysis of contemporary culture and social aspect in mid-20th century. This study is to explore the diversity of animation representation, aesthetics, and also to suggest a new aspect of animation studies.

• Cartoon and Animation Studies
• /
• s.44
• /
• pp.117-141
• /
• 2016

The various representations are receiving attention in modern society that has so many contents. Among them, the fantasy shows that you can not see in reality. But the intention of these fantasies is not giving a visual fun. The fantasy show the reality through stories that are not in reality. The fantasy that allows readers to continue to make the suspect between the real and the imagination and that suspect arises from the desire of real life. If the desire break the community, the social ideology will collapses. Conversely, if the desire is overturned by community, the social will be maintained. The goal of the fantasy which has the relationships of society is revealed through the various expressions of existence. They are divided into the subject and the other show the inner side of the main character. The subject shows the inner side of the main character by the transformation, alter ego and the other exists. The other shows the desire of the subject by the transformation, alter ego and the strangers. Disney animation studios select the target audience and the message in relation with the society. They choose the original like the fairy tale, myth and change them to satisfy the middle class. The characters of Disney animation says that messages by the expression of fantasy. The subject go through the transformation by twice. The first transformation is antisocial and the second transformation is social. The second shows a complete transformation. The other characters personified show the many kinds of the main character. The other as the alter ego of the main character represents the desire of the subject. They are described as an object of fear and exclusion. They expresses as a dark and menacing looks and hinders the complete transformation of the subject. But they were overthrown by the subject at the end of the story and it strengthen the social ideology. As a result, Disney highlights the value and the moral message of the society by using the representation of fantasy.

• Journal of Intelligence and Information Systems
• /
• v.18 no.4
• /
• pp.97-116
• /
• 2012

Conceptual modeling is an important step for successful system development. It helps system designers and business practitioners share the same view on domain knowledge. If the work is successful, a result of conceptual modeling can be beneficial in increasing productivity and reducing failures. However, the value of conceptual modeling is unlikely to be evaluated uniformly because we are lack of agreement on how to elicit concepts and how to represent those with conceptual modeling constructs. Especially, designing relationships between components, also known as part-whole relationships, have been regarded as complicated work. The recent study, "Representing Part-Whole Relations in Conceptual Modeling : An Empirical Evaluation" (Shanks et al., 2008), published in MIS Quarterly, can be regarded as one of positive efforts. Not only the study is one of few attempts of trying to clarify how to select modeling alternatives in part-whole design, but also it shows results based on an empirical experiment. Shanks et al. argue that there are two modeling alternatives to represent part-whole relationships : an implicit representation and an explicit one. By conducting an experiment, they insist that the explicit representation increases the value of a conceptual model. Moreover, Shanks et al. justify their findings by citing the BWW ontology. Recently, the study from Shanks et al. faces criticism. Allen and March (2012) argue that Shanks et al.'s experiment is lack of validity and reliability since the experimental setting suffers from error-prone and self-defensive design. They point out that the experiment is intentionally fabricated to support the idea, as such that using concrete UML concepts results in positive results in understanding models. Additionally, Allen and March add that the experiment failed to consider boundary conditions; thus reducing credibility. Shanks and Weber (2012) contradict flatly the argument suggested by Allen and March (2012). To defend, they posit the BWW ontology is righteously applied in supporting the research. Moreover, the experiment, they insist, can be fairly acceptable. Therefore, Shanks and Weber argue that Allen and March distort the true value of Shanks et al. by pointing out minor limitations. In this study, we try to investigate the dispute around Shanks et al. in order to answer to the following question : "What is the proper value of the study conducted by Shanks et al.?" More profoundly, we question whether or not using the BWW ontology can be the only viable option of exploring better conceptual modeling methods and procedures. To understand key issues around the dispute, first we reviewed previous studies relating to the BWW ontology. We critically reviewed both of Shanks and Weber and Allen and March. With those findings, we further discuss theories on part-whole (or part-of) relationships that are rarely treated in the dispute. As a result, we found three additional evidences that are not sufficiently covered by the dispute. The main focus of the dispute is on the errors of experimental methods: Shanks et al. did not use Bunge's Ontology properly; the refutation of a paradigm shift is lack of concrete, logical rationale; the conceptualization on part-whole relations should be reformed. Conclusively, Allen and March indicate properly issues that weaken the value of Shanks et al. In general, their criticism is reasonable; however, they do not provide sufficient answers how to anchor future studies on part-whole relationships. We argue that the use of the BWW ontology should be rigorously evaluated by its original philosophical rationales surrounding part-whole existence. Moreover, conceptual modeling on the part-whole phenomena should be investigated with more plentiful lens of alternative theories. The criticism on Shanks et al. should not be regarded as a contradiction on evaluating modeling methods of alternative part-whole representations. To the contrary, it should be viewed as a call for research on usable and useful approaches to increase value of conceptual modeling.

• Journal of the Korean Institute of Landscape Architecture
• /
• v.45 no.3
• /
• pp.66-79
• /
• 2017

In 2014, a Donghak Peasant Revolution Memorial Park design competition was held with various forms and techniques to convey mourning. This is a process of the reconsideration of memorial projects that are used to stimulate the collective memory and it is a meaningful resource for examining the consciousness of contemporary designers in regards to the memorial designs that are currently under planning in Korea. This study investigated the background of the Donghak commemorative projects that took place at the same site in a timely manner and analyzed the design competition through the existing literature research. Through this, it was seen that the memorial, which was formed by means of past political purposes, has changed into a way to collect various opinions and forms through open design competition. A framework of analysis prepared through multi-layer analysis is daily use, interaction and spontaneity, abstraction, temporality, locality, integration and harmony with surroundings. The results of this study are as follows. First, in order to convey memorial commemoration in everyday life, the projects organized scattered memorial spaces with special characteristics and linked them with daily activities program. Second, the projects used direct participation and emotional experiences to interact with monuments. Third, color, vertical elements, clustering, and park frame manipulation were used for abstract reproduction. Fourth, the projects introduce architecture and furniture that can be changed and plants for temporal change. Fifth, the previous terrain was restored and the setting of the scene was reproduced in order to make the site a space with place. Sixth, to improve the connection with existing monuments, the projects used techniques such as relaxation and the reinforcement of circulation lines and axes. Seventh, a path and a building conforming to the terrain were arranged for harmony with the surroundings.