• Title/Summary/Keyword: Generate Data

Search Result 3,066, Processing Time 0.03 seconds

Bit-width Aware Generator and Intermediate Layer Knowledge Distillation using Channel-wise Attention for Generative Data-Free Quantization

  • Jae-Yong Baek;Du-Hwan Hur;Deok-Woong Kim;Yong-Sang Yoo;Hyuk-Jin Shin;Dae-Hyeon Park;Seung-Hwan Bae
    • 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.

Automatic scoring of mathematics descriptive assessment using random forest algorithm (랜덤 포레스트 알고리즘을 활용한 수학 서술형 자동 채점)

  • Inyong Choi;Hwa Kyung Kim;In Woo Chung;Min Ho Song
    • The Mathematical Education
    • /
    • v.63 no.2
    • /
    • pp.165-186
    • /
    • 2024
  • Despite the growing attention on artificial intelligence-based automated scoring technology as a support method for the introduction of descriptive items in school environments and large-scale assessments, there is a noticeable lack of foundational research in mathematics compared to other subjects. This study developed an automated scoring model for two descriptive items in first-year middle school mathematics using the Random Forest algorithm, evaluated its performance, and explored ways to enhance this performance. The accuracy of the final models for the two items was found to be between 0.95 to 1.00 and 0.73 to 0.89, respectively, which is relatively high compared to automated scoring models in other subjects. We discovered that the strategic selection of the number of evaluation categories, taking into account the amount of data, is crucial for the effective development and performance of automated scoring models. Additionally, text preprocessing by mathematics education experts proved effective in improving both the performance and interpretability of the automated scoring model. Selecting a vectorization method that matches the characteristics of the items and data was identified as one way to enhance model performance. Furthermore, we confirmed that oversampling is a useful method to supplement performance in situations where practical limitations hinder balanced data collection. To enhance educational utility, further research is needed on how to utilize feature importance derived from the Random Forest-based automated scoring model to generate useful information for teaching and learning, such as feedback. This study is significant as foundational research in the field of mathematics descriptive automatic scoring, and there is a need for various subsequent studies through close collaboration between AI experts and math education experts.

A Basic Study on User Experience Evaluation Based on User Experience Hierarchy Using ChatGPT 4.0 (챗지피티 4.0을 활용한 사용자 경험 계층 기반 사용자 경험 평가에 관한 기초적 연구)

  • Soomin Han;Jae Wan Park
    • The Journal of the Convergence on Culture Technology
    • /
    • v.10 no.2
    • /
    • pp.493-498
    • /
    • 2024
  • With the rapid advancement of generative artificial intelligence technology, there is growing interest in how to utilize it in practical applications. Additionally, the importance of prompt engineering to generate results that meet user demands is being newly highlighted. Exploring the new possibilities of generative AI can hold significant value. This study aims to utilize ChatGPT 4.0, a leading generative AI, to propose an effective method for evaluating user experience through the analysis of online customer review data. The user experience evaluation method was based on the six-layer elements of user experience: 'functionality', 'reliability', 'usability', 'convenience', 'emotion', and 'significance'. For this study, a literature review was conducted to enhance the understanding of prompt engineering and to grasp the clear concept of the user experience hierarchy. Based on this, prompts were crafted, and experiments for the user experience evaluation method were carried out using the analysis of collected online customer review data. In this study, we reveal that when provided with accurate definitions and descriptions of the classification processes for user experience factors, ChatGPT demonstrated excellent performance in evaluating user experience. However, it was also found that due to time constraints, there were limitations in analyzing large volumes of data. By introducing and proposing a method to utilize ChatGPT 4.0 for user experience evaluation, we expect to contribute to the advancement of the UX field.

Factors Influencing Effects of Korea's Rural Life Improvement Program on Quality of Life of Rural Women (한국의 농촌 생활개선사업이 농촌여성의 삶의 질에 미치는 영향요인)

  • Bereket Roba Gamo;Yoon-Ji Choi;Jung-Shin Choi;Joo-Lee Son
    • Journal of Agricultural Extension & Community Development
    • /
    • v.30 no.4
    • /
    • pp.243-257
    • /
    • 2023
  • Rural life improvement programs (RLIPs) have been implemented with a central goal of improving the quality of rural life and promoting rural welfare and cultural life. However, different factors may influence the effect of rural life improvement programs on rural communities or households. This study aimed to investigate the determinants of perceived effects of RLIPs on quality of life of rural women in South Korea. We used a mixed research design to generate data for this study. We collected survey data from 311 rural women who participated in the RLIPs and also conducted a focus group discussion. We analyzed the quantitative data using descriptive statistics and hierarchical regression to identify the variables that predicted effects of RLIPs on quality of life of rural women. Our study finds that age, type of residence in the community, leadership experience, level of education, community satisfaction and community participation influenced respondents' perceived effects of RLIPs. The results imply that the benefits of a development intervention could not be uniformly reaped by residents of a community.

LNG Gas Demand Forecasting in Incheon Port based on Data: Comparing Time Series Analysis and Artificial Neural Network (데이터 기반 인천항 LNG 수요예측 모형 개발: 시계열분석 및 인공신경망 모형 비교연구)

  • Beom-Soo Kim;Kwang-Sup Shin
    • The Journal of Bigdata
    • /
    • v.8 no.2
    • /
    • pp.165-175
    • /
    • 2023
  • LNG is a representative imported cargo at Incheon Port and has a relatively high contribution to the increase/decrease in overall cargo volume at Incheon Port. In addition, in the view point of nationwide, LNG is the one of the most important key resource to supply the gas and generate electricity. Thus, it is very essential to identify the factors that have impact on the demand fluctuation and build the appropriate forecasting model, which present the basic information to make balance between supply and demand of LNG and establish the plan for power generation. In this study, different to previous research based on macroscopic annual data, the weekly demand of LNG is converted from the cargo volume unloaded by LNG carriers. We have identified the periodicity and correlations among internal and external factors of demand variability. We have identified the input factors for predicting the LNG demand such as seasonality of weekly cargo volume, the peak power demand, and the reserved capacity of power supply. In addition, in order to predict LNG demand, considering the characteristics of the data, time series prediction with weekly LNG cargo volume as a dependent variable and prediction through an artificial neural network model were made, the suitability of the predictions was verified, and the optimal model was established through error comparison between performance and estimates.

Privilege and Immunity of Information and Data from Aviation Safety Program in Unites States (미국 항공안전데이터 프로그램의 비공개 특권과 제재 면제에 관한 연구)

  • Moon, Joon-Jo
    • The Korean Journal of Air & Space Law and Policy
    • /
    • v.23 no.2
    • /
    • pp.137-172
    • /
    • 2008
  • The earliest safety data programs, the FDR and CVR, were electronic reporting systems that generate data "automatically." The FDR program, originally instituted in 1958, had no publicly available restrictions for protections against sanctions by the FAA or an airline, although there are agreements and union contracts forbidding the use of FDR data for FAA enforcement actions. This FDR program still has the least formalized protections. With the advent of the CVR program in 1966, the precursor to the current FAR 91.25 was already in place, having been promulgated in 1964. It stated that the FAA would not use CVR data for enforcement actions. In 1982, Congress began restricting the disclosure of the CVR tape and transcripts. Congress added further clarification of the availability of discovery in civil litigation in 1994. Thus, the CVR data have more definitive protections in place than do FDR data. The ASRS was the first non-automatic reporting system; and built into its original design in 1975 was a promise of limited protection from enforcement sanctions. That promise was further codified in an FAR in 1979. As with the CVR, from its inception, the ASRS had some protections built in for the person who might have had a safety problem. However, the program did not (and to this day does not) explicitly deal with issues of use by airlines, litigants, or the public media, although it appears that airlines will either take a non-punitive stance if an ASRS report is filed, or the airline may ignore the fact that it has been filed at all. The FAA worked with several U.S. airlines in the early 1990s on developing ASAP programs, and the FAA issued an Advisory Circular about the program in 1997. From its inception, the ASAP program contained some FAA enforcement protections and company discipline protections, although some protection against litigation disclosure and public disclosure was not added until 2003, when FAA Order 8000.82 was promulgated, placing the program under the protections of FAR 193, which had been added in 2001. The FOQA program, when it was first instituted through a demonstration program in 1995, did not contain protections against sanctions. Now, however, the FAA cannot take enforcement action based on FOQA safety data, and an airline is limited to "corrective action" under the program. Union contracts can exclude FOQA from the realm of disciplinary action, although airline practice may be for airlines to require retraining if there is no contract in place forbidding it. The data is protected against disclosure for litigation and public media purposes by FAA Order 8000.81, issued in 2003, which placed FOQA under the protections of FAR 193. The figure on the next page shows when each program began, and when each statute, regulation, or order became effective for that program.

  • PDF

A Test Stream Generating Method for the Digital TV Software (디지털 TV 소프트웨어를 위한 테스트 스트림 자동 생성 방법)

  • 곽태희;최병주
    • Journal of KIISE:Software and Applications
    • /
    • v.30 no.10
    • /
    • pp.925-937
    • /
    • 2003
  • The input of digital TV software is the Transport Stream, which utilizes the moving picture compression technique, MPEC-2 (Moving Picture Experts Groups-2). MPEG-2 TS consists of the complicated table hierarchy and internal relationships, as well as the various restrictions with regards to system standard of digital TV software in determining the field values of transport stream. However, the general MPEG-2 TS generation tool produces transport streams solely based on the MPEG-2 TS specification itself, and does not consider the interaction between modulo features or modules themselves, which construct digital TV software. In this paper, we propose a method to systematically generate MPEG-2 TS test data, namely‘Test Stream’, for digital TV software. We present the experiment result where the test stream derived from our developed tool according to the proposed method was applied to the actual digital TV software settop-box, and then analyze the result. Apart from other existing MPEG-2 TS generation tools, the advantage of our method is that not only is it capable of generating various levels of test streams including digital TV software module, integration, and system testing, but also detecting errors and analyzing their causes.

Annotation and Expression Profile Analysis of cDNAs from the Antarctic Diatom Chaetoceros neogracile

  • Jung, Gyeong-Seo;Lee, Choul-Gyun;Kang, Sung-Ho;Jin, Eon-Seon
    • Journal of Microbiology and Biotechnology
    • /
    • v.17 no.8
    • /
    • pp.1330-1337
    • /
    • 2007
  • To better understand the gene expression of the cold-adapted polar diatom, we conducted a survey of the Chaetoceros neogracile transcriptome by cDNA sequencing and expression of interested cDNAs from the Antarctic diatom. A non-normalized cDNA library was constructed from the C. neogracile, and a total of 2,500 cDNAs were sequenced to generate 1,881 high-quality expressed sequence tags (ESTs) (accession numbers EL620615-EL622495). Based on their clustering, we identified 154 unique clusters comprising 342 ESTs. The remaining 1,540 ESTs did not cluster. The number of unique genes identified in the data set is thus estimated to be 1,694. Taking advantage of various tools and databases, putative functions were assigned to 939 (55.4%) of these genes. Of the remaining 540 (31.9%) unknown sequences, 215 (12.7%) appeared to be C. neogracile-specific since they lacked any significant sequence similarity to any sequence available in the public databases. C. neogracile consisted of a relatively high percentage of genes involved in metabolism, genetic information processing, cellular processes, defense or stress resistance, photosynthesis, structure, and signal transduction. From the ESTs, the expression of these putative C. neogracile genes was investigated: fucoxanthin chlorophyll (chl) a,c-binding protein (FCP), ascorbate peroxidase (ASP), and heat-shock protein 90 (HSP90). The abundance of ASP and HSP90 changed substantially in response to different culture conditions, indicating the possible regulation of these genes in C. neogracile.

Stereo-To-Multiview Conversion System Using FPGA and GPU Device (FPGA와 GPU를 이용한 스테레오/다시점 변환 시스템)

  • Shin, Hong-Chang;Lee, Jinwhan;Lee, Gwangsoon;Hur, Namho
    • Journal of Broadcast Engineering
    • /
    • v.19 no.5
    • /
    • pp.616-626
    • /
    • 2014
  • In this paper, we introduce a real-time stereo-to-multiview conversion system using FPGA and GPU. The system is based on two different devices so that it consists of two major blocks. The first block is a disparity estimation block that is implemented on FPGA. In this block, each disparity map of stereoscopic video is estimated by DP(dynamic programming)-based stereo matching. And then the estimated disparity maps are refined by post-processing. The refined disparity map is transferred to the GPU device through USB 3.0 and PCI-express interfaces. Stereoscopic video is also transferred to the GPU device. These data are used to render arbitrary number of virtual views in next block. In the second block, disparity-based view interpolation is performed to generate virtual multi-view video. As a final step, all generated views have to be re-arranged into a single image at full resolution for presenting on the target autostereoscopic 3D display. All these steps of the second block are performed in parallel on the GPU device.

Accuracy Analysis of DEMs Generated from High Resolution Optical and SAR Images (고해상도 광학영상과 SAR영상으로부터 생성된 수치표고모델의 정확도 분석)

  • Kim, Chung;Lee, Dong-Cheon;Yom, Jae-Hong;Lee, Young-Wook
    • Proceedings of the Korean Society of Surveying, Geodesy, Photogrammetry, and Cartography Conference
    • /
    • 2004.04a
    • /
    • pp.337-343
    • /
    • 2004
  • Spatial information could be obtained from spaceborne high resolution optical and synthetic aperture radar(SAR) images. However, some satellite images do not provide physical sensor information instead, rational polynomial coefficients(RPC) are available. The objectives of this study are: (1) 3-dimensional ground coordinates were computed by applying rational function model(RFM) with the RPC for the stereo pair of Ikonos images and their accuracy was evaluated. (2) Interferometric SAR(InSAR) was applied to JERS-1 images to generate DEM and its accuracy was analysis. (3) Quality of the DEM generated automatically also analyzed for different types of terrain in the study site. The overall accuracy was evaluated by comparing with GPS surveying data. The height offset in the RPC was corrected by estimating bias. In consequence, the accuracy was improved. Accuracy of the DEMs generated from InSAR with different selection of GCP was analyzed. In case of the Ikonos images, the results show that the overall RMSE was 0.23327", 0.l1625" and 13.70m in latitude, longitude and height, respectively. The height accuracy was improved after correcting the height offset in the RPC. i.e., RMSE of the height was 1.02m. As for the SAR image, RMSE of the height was 10.50m with optimal selection of GCP. For the different terrain types, the RMSE of the height for urban, forest and flat area was 23.65m, 8.54m, 0.99m, respectively for Ikonos image while the corresponding RMSE was 13.82m, 18.34m, 10.88m, respectively lot SAR image.

  • PDF