• Proceedings of the Korea Database Society Conference
• /
• 2001.06a
• /
• pp.331-341
• /
• 2001

본 논문에서는 Naive Bayes 문서 분류기를 위한 새로운 학습모델을 제안한다. 이 모델에서는 라벨이 없는 문서들의 집합으로부터 선택한 적은 수의 학습 문서들을 이용하여 문서 분류기를 재학습한다. 본 논문에서는 이러한 학습 방법을 따를 경우 작은 비용으로도 문서 분류기의 정확도가 크게 향상될 수 있다는 사실을 보인다. 이와 같이, 알고리즘을 통해 라벨이 없는 문서들의 집합으로부터 정보량이 큰 문서를 선택한 후, 전문가가 이 문서에 라벨을 부여하는 방식으로 학습문서를 결정하는 것을 selective sampling이라 한다. 본 논문에서는 이러한 selective sampling 문제를 Naive Bayes 문서 분류기에 적용한다. 제안한 학습 방법에서는 라벨이 없는 문서들의 집합으로부터 재학습 문서를 선택하는 기준 측정치로서 평균절대편차(Mean Absolute Deviation), 엔트로피 측정치를 사용한다. 실험을 통해서 제안한 학습 방법이 기존의 방법인 신뢰도(Confidence measure)를 이용한 학습 방법보다 Naive Bayes 문서 분류기의 성능을 더 많이 향상시킨다는 사실을 보인다.

• Transactions on Semiconductor Engineering
• /
• v.2 no.4
• /
• pp.46-51
• /
• 2024

Deep Reinforcement Learning (DRL) has demonstrated human-level performance in sequential decision-making tasks and enables edge devices to adapt autonomously to unknown environments. However, implementing DRL adaptation remains challenging due to its massive data interactions and extensive DNN computations. Existing FPGA-based DRL accelerators focus solely on computation acceleration, leading to prolonged adaptation times. This paper proposes an energy-efficient FPGA accelerator tailored for fast online DRL adaptation, leveraging three key innovations: 1) A Heterogeneous Replay Buffer (HRB) that reduces training iterations by up to 90%, 2) Mixed-Precision Selective Re-Training (MP-SELRET) that decreases computations by 12% while replacing 27.2% of 32-bit floating-point operations with 16-bit fixed-point operations, 3) A Mixed-Precision Heterogeneous Architecture (MPHA) that maximizes resource utilization and boosts throughput by 39.8%. The proposed accelerator significantly enhances the efficiency and speed of DRL adaptation, addressing the limitations of traditional scratch trainingmethods.

• KIPS Transactions on Software and Data Engineering
• /
• v.10 no.9
• /
• pp.349-358
• /
• 2021

Recently, a method of finding the attention area or localization area for an object of an image using CAM (Class Activation Map)[1] has been variously carried out as a study of WSOL (Weakly Supervised Object Localization). The attention area extraction from the object heat map using CAM has a disadvantage in that it cannot find the entire area of the object by focusing mainly on the part where the features are most concentrated in the object. To improve this, using CAM and Selective Search[6] together, we first expand the attention area in the heat map, and a Gaussian smoothing is applied to the extended area to generate retraining data. Finally we train the data to expand the attention area of the objects. The proposed method requires retraining only once, and the search time to find an localization area is greatly reduced since the selective search is not needed in this stage. Through the experiment, the attention area was expanded from the existing CAM heat maps, and in the calculation of IOU (Intersection of Union) with the ground truth for the bounding box of the expanded attention area, about 58% was improved compared to the existing CAM.

• The Journal of the Korea Contents Association
• /
• v.13 no.2
• /
• pp.8-18
• /
• 2013

In this paper, we capture learning while taking a video, play time and time line of the video frame in the form of areas that require re-learning in HTML5 mobile web store. When you select an image frame can display a list of the frame to take advantage of HTML5 Video tag up to 9 capture and save the playing time at the position. Implemented in a manner that runs Effects as compared to learning to run the entire frame capture learning and re-learning frame partial immersion learners matchumhyeong storytelling can be implemented. Interval Iterative Learning in a random order, so learners can level alignment by iterative learning on academic performance can have a positive effect.

• Cartoon and Animation Studies
• /
• s.32
• /
• pp.363-381
• /
• 2013

In this paper, on the web, offline mobile learning content to reinforce the learning of the video frame-by-frame necessary for re-learning area to capture only the important areas. The frame of the captured image and the image in the form of advanced training time saved and also a description of the notes feature to store. The area needed for the capture area re-learning the learner to learner-centered custom systems can be applied. In order to capture the learning program, regardless of the configuration of the selected frame by frame in order to capture the user-centric storytelling-based learning can be applied. Capture the full effect of the system compared to learning and learner-centered learning time-saving reconstruction of the frame according to the customized learning to play a positive role in improving effectiveness.

• Proceedings of the KIEE Conference
• /
• 2000.07a
• /
• pp.102-106
• /
• 2000

본 논문은 뉴로-퍼지 모델의 구조학습을 이용하여 한 시간 앞의 전력 수요를 예측하는 체계적인 방법을 제안한다. 제안된 예측시스템은 시간 단위로 뉴로-퍼지 모델을 재학습하기 위해서 필요한 초기 구조를 요일 유형과 시간 별로 미리 생성하고, 이를 초기 구조 뱅크에 저장한다. 예측이 수행되는 시점의 요일 유형에 따라 선택된 초기 구조를 이용하여 뉴로-퍼지 모델을 초기화하고, 학습하고, 예측을 수행한다. 제안된 방법의 실효성을 검증하기 위해 1996년과 1997년의 실제 전력 수요 데이터를 이용하여 모의 실험을 수행한다. 실험결과 제안된 방법은 기존의 다층 퍼셉트론을 이용한 방법과 비교하여 예측의 정확도 측면과 신뢰도 측면에서 모두 향상된 결과를 얻는다.

• The Journal of Korean Association of Computer Education
• /
• v.3 no.1
• /
• pp.51-63
• /
• 2000

All learners can selection of their work through hypermedia technology in the area provided by usual WBI. Also, it provides learner with individual teaching-learning environment and estimation. RC2 System has the fundamental client/server model, and provides the learning, evaluation algorithms based on the LCPG(Learning Contents Problem Graph) model, the dynamic re-learning mechanism in according to the property of individual. Moreover, it support learning editor to provide interface, which is convenient for teacher, Courseware writer, on the Web

• Korean Journal of Cognitive Science
• /
• v.31 no.4
• /
• pp.137-154
• /
• 2020

Psychological properties of learners have influence on learning behaviors in various ways. The purpose of this study was to examine how the goal orientation of learners affected the learning time distribution method. Regulatory focus and theories of intelligence were measured and manipulated in order to differentiate participants' goal-oriented state. Two variables are known to be key variables influencing learner's goal orientation, inducing the approach-avoidance strategy and mastery-performance oriented attitude. In the experiment, the control focus was divided into two groups based on the inclination test score (regulatory Focus Questionnaire, RFQ), and TOI(theory of intelligence) was temporally induced through manipulation to confirm the interaction between the two variables. Participants were able to determine the order of learning freely by learning a set of Spanish-Korean word pairs and then selecting the items they would like to re-learn. Word pairs consisted of difficult or easy items, and learners could learn the same word many times if they wanted to. In the results, promotion-incremental group showed allocating difficult word-pairs in early time.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2013.10a
• /
• pp.496-498
• /
• 2013

Smartphone is becoming popular from younger generation to old generation because of its convenience. Generalizing use of smartphone, it was being developed and released for education application from based on PC to smartphone. in this paper, this application was designed and developed an helpful and enjoyable application for those who very interested child education. in this paper, suggestion of the propose of education is provided on service various contetns. The children can do implement information about organism characters living in sea just selected them they want without going to sea and aquarium whenever they want. this application is implemented that children also can be relearning the organism characters by quiz that already learned them before, and developing artistic sense, creativity and imagination by coloring exercise.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.20 no.1
• /
• pp.7-14
• /
• 2010

In this paper, we introduce a hierarchical internet application traffic classification system based on SVM as an alternative overcoming the uppermost limit of the conventional methodology which is using the port number or payload information. After selecting an optimal attribute subset of the bidirectional traffic flow data collected from the campus, the proposed system classifies the internet application traffic hierarchically. The system is composed of three layers: the first layer quickly determines P2P traffic and non-P2P traffic using a SVM, the second layer classifies P2P traffics into file-sharing, messenger, and TV, based on three SVDDs. The third layer makes specific classification of the entire 16 application traffics. By classifying the internet application traffic finely or coarsely, the proposed system can guarantee an efficient system resource management, a stable network environment, a seamless bandwidth, and an appropriate QoS. Also, even a new application traffic is added, it is possible to have a system incremental updating and scalability by training only a new SVDD without retraining the whole system. We validate the performance of our approach with computer experiments.