• Journal of Intelligence and Information Systems
• /
• v.14 no.3
• /
• pp.59-76
• /
• 2008

In this paper, we propose an agent architecture called L-CAA that is quite effective in real-time dynamic environments. L-CAA is an extension of CAA, the behavior-based agent architecture which was also developed by our research group. In order to improve adaptability to the changing environment, it is extended by adding reinforcement learning capability. To obtain stable performance, however, behavior selection and execution in the L-CAA architecture do not entirely rely on learning. In L-CAA, learning is utilized merely as a complimentary means for behavior selection and execution. Behavior selection mechanism in this architecture consists of two phases. In the first phase, the behaviors are extracted from the behavior library by checking the user-defined applicable conditions and utility of each behavior. If multiple behaviors are extracted in the first phase, the single behavior is selected to execute in the help of reinforcement learning in the second phase. That is, the behavior with the highest expected reward is selected by comparing Q values of individual behaviors updated through reinforcement learning. L-CAA can monitor the maintainable conditions of the executing behavior and stop immediately the behavior when some of the conditions fail due to dynamic change of the environment. Additionally, L-CAA can suspend and then resume the current behavior whenever it encounters a higher utility behavior. In order to analyze effectiveness of the L-CAA architecture, we implement an L-CAA-enabled agent autonomously playing in an Unreal Tournament game that is a well-known dynamic virtual environment, and then conduct several experiments using it.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.7
• /
• pp.11-20
• /
• 2024

In this paper, we propose the BAG (Bit-width Aware Generator) and the Intermediate Layer Knowledge Distillation using Channel-wise Attention to reduce the knowledge gap between a quantized network, a full-precision network, and a generator in GDFQ (Generative Data-Free Quantization). Since the generator in GDFQ is only trained by the feedback from the full-precision network, the gap resulting in decreased capability due to low bit-width of the quantized network has no effect on training the generator. To alleviate this problem, BAG is quantized with same bit-width of the quantized network, and it can generate synthetic images, which are effectively used for training the quantized network. Typically, the knowledge gap between the quantized network and the full-precision network is also important. To resolve this, we compute channel-wise attention of outputs of convolutional layers, and minimize the loss function as the distance of them. As the result, the quantized network can learn which channels to focus on more from mimicking the full-precision network. To prove the efficiency of proposed methods, we quantize the network trained on CIFAR-100 with 3 bit-width weights and activations, and train it and the generator with our method. As the result, we achieve 56.14% Top-1 Accuracy and increase 3.4% higher accuracy compared to our baseline AdaDFQ.

• The Journal of Small Business Innovation
• /
• v.19 no.2
• /
• pp.53-68
• /
• 2016

Motivation and activities for technological learning, entrepreneurship, innovation, and creativity are driving forces of economic development in Asian countries. In the early stages of technological development, technological learning and entrepreneurship are efficient ways in which to catch up with advanced countries because firms can accumulate skills and knowledge quickly at relatively low risk. In the later stages of technological development, however, innovation and creativity become more important. This study aims to identify a) the factors (learning capabilities) that influence technological learning performance and b) barriers to enhancing innovation capabilities for the creative economy and organizations. The major part of this study is related to learning capabilities in the post-catch-up era. Based on a literature review and observations from Korean experiences, this study proposes a technological learning model composed of various influencing factors on technological learning. Three hypotheses are derived, and data are collected from Korean machine tool manufacturers. Intense interviews with CEOs and R&D directors are conducted using structured questionnaires. Statistical analysis, such as correlation and ANOVA are then carried out. Furthermore, this study addresses how to enhance innovation capabilities to move forward. Innovation enablers and barriers are identified by case studies and policy analysis. The results of the empirical study identify several levels of firms' learning capabilities and activities such as a) stock of technology, b) potential of technical labor, c) explicit technological efforts, d) readiness to learn, e) top management support, f) a formal technological learning system, g) high learning motivation, h) appropriate technology choice, and i) specific goal setting. These learning capabilities determine firms' learning performance, especially in the early stages of development. Furthermore, it is found that the critical factors for successful technological learning vary along the stages of technology development. Throughout the statistical and policy analyses, this study confirms that technological learning can be understood as an intrinsic principle of the technology development process. Firms perform proactive and creative learning in the late stages, while reactive and imitative learning prevails in the early stages. In addition, this study identifies the driving forces or facilitating factors enhancing innovation performance in the post catch-up era. The results of the preliminary case studies and policy analysis show some facilitating factors such as a) the strategic intent of the CEO and corporate culture, b) leadership and change agents, c) design principles and routines, d) ecosystem and collaboration with partners, and e) intensive R&D investment.

• Journal of The Korean Association of Information Education
• /
• v.12 no.3
• /
• pp.275-282
• /
• 2008

If a learner should fail to complete previous courses before moving on to the next level will face difficulty keeping up with it. Such personal tuition for those having trouble coping with their class is an issue that needs urgent addressing, which cannot be a burden only to teachers, this study has been conducted to sought for a solution. This study has developed and put into application a web based analyzing system to assess the area of deficit in students followed by obliterating accumulated learning deficits to impart assistance for their study. The subject of the study comprised of ten students from a school where the researcher is on duty and the field of assessment with the analyzing system were numbers and calculations. As a result, we could find out its efficiency in elevating their capability and interest in learning, which was proven to be statistically significant using ANOVA.

• Journal of the Korean Data and Information Science Society
• /
• v.27 no.2
• /
• pp.327-335
• /
• 2016

The People's Republic of China has vigorously been pursuing the internationalization of the Chinese Yuan or Renminbi after the financial crisis of 2008. In this view, an abrupt increase of use of the Chinese Yuan in the onshore and offshore markets are important milestones to be one of important currencies. One of the most frequently used methods to forecast volatility is GARCH model. Since a prediction error of the GARCH model has been reported quite high, a lot of efforts have been made to improve forecasting capability of the GARCH model. In this paper, we have proposed MLP-GARCH and a DL-GARCH by employing Artificial Neural Network to the GARCH. In an application to forecasting Chinese Yuan volatility, we have successfully shown their overall outperformance in forecasting over the GARCH.

• Journal of The Korean Association of Information Education
• /
• v.16 no.2
• /
• pp.211-222
• /
• 2012

21C future learners are required to have creative thinking and problem-solving capability to address an issue wisely by integrating and applying knowledge and functions. The robot education that has recently been studied in primary and secondary schools was reported to be effective in satisfying the requirements. The robot education varies ranging from the existing after-school education to an integrated approach used for regular curriculums, and has actively been studied. Nevertheless, aside from positive study results, any studies on the environment where primary school students can learn robot and programming knowledge more friendly is still insufficient. Therefore, this study was intended to give students a robot class with the use of storytelling friendly to students in order for primary school students to learn robot and programming knowledge with ease. The study result showed that acquirement of programming knowledge was improved, and that the students had a positive learning attitude. In addition, it was found that the storytelling of the robot class helped provide the entire learning context and continuous learning motivation for the students.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.26 no.6
• /
• pp.108-114
• /
• 1989

Introducing and optimizing bit-significance to the Hopfield model, ten highly correlated binary images, i.e., numbers "0" to "9", are successfully stored and retrieved in a $6{}8$ node system. Unlike many other neural network models, this model has stronger error correction capability for correlated images such as "6","8","3", and "9". The bit significance optimization is regarded as an adaptive learning process based on least-mean-square error algorithm, and may be implemented with Widrow-Hoff neural nets optimizer. A design for electro-optic implementation including the adaptive optimization networks is also introduced.

• The Journal of Korean Association of Computer Education
• /
• v.15 no.1
• /
• pp.33-41
• /
• 2012

The purpose of this study was to examine the effectiveness of problem-based learning(PBL) of information communication ethics on the problem-solving skills of fifth- and sixth-grade elementary school students about information communication ethics and their attitude. The findings of the study were as follows: First, the PBL-based teaching materials for information communication ethics could serve to stir up the interest of students by letting them face problem situations caused by the disfunction of the information society. Second, a wide variety of teaching-learning activities including cooperative learning and discussion were implemented while the selected students performed the given problem-solving tasks, and the activities had a very good effect on PBL-based information communication ethics education. Third, PBL-based information communication ethics education could be one of effective ways to boost not only the problem-solving skills of students but the teaching interest and capability of teachers.

• Journal of Korea Multimedia Society
• /
• v.23 no.4
• /
• pp.525-538
• /
• 2020

A practical animal face classification system that classifies animals in image and video data is considered as a pivotal topic in machine learning. In this research, we are proposing a novel method of fully connected dual Deep Convolutional Neural Network (DCNN), which extracts and analyzes image features on a large scale. With the inclusion of the state of the art Batch Normalization layer and Exponential Linear Unit (ELU) layer, our proposed DCNN has gained the capability of analyzing a large amount of dataset as well as extracting more features than before. For this research, we have built our dataset containing ten thousand animal faces of ten animal classes and a dual DCNN. The significance of our network is that it has four sets of convolutional functions that work laterally with each other. We used a relatively small amount of batch size and a large number of iteration to mitigate overfitting during the training session. We have also used image augmentation to vary the shapes of the training images for the better learning process. The results demonstrate that, with an accuracy rate of 92.0%, the proposed DCNN outruns its counterparts while causing less computing costs.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.9 no.6
• /
• pp.650-656
• /
• 1999

In this paper, we study on the convergence of the evolution strategies by introducing the Lamarckian evolution and the Baldwin effect, and propose a random local searching and a reinforcement local searching methods. In the random local searching method some neighbors generated randomly from each individual are med without any other information, but in the reinforcement local searching method the previous results of the local search are reflected on the current local search. From the viewpoint of the purpose of the local search it is suitable that we try all the neighbors of the best individual and then search the neighbors of the best one of them repeatedly. Since the reinforcement local searching method based on the Lamarckian evolution and Baldwin effect does not search neighbors randomly, but searches the neighbors in the direction of the better fitness, it has advantages of fast convergence and an improvement on the global searching capability. In other words the performance of the evolution strategies is improved by introducing the learning, reinforcement local search, into the evolution. We study on the learning effect on evolution strategies by applying the proposed method to various function optimization problems.