• Annual Conference of KIPS
• /
• 2024.10a
• /
• pp.40-41
• /
• 2024

최근 딥러닝 연산의 고도화에 따라 희소성(Sparsity)을 효율적으로 처리할 수 있는 유연한 구조의 DNN 가속기가 중요해지고 있다. 그러나 기존의 가속기들은 유연성과 효율성 면에서 한계가 존재한다. 본 논문에서는 DNN 가속기의 기존 모델들과 최신 연구 동향에 대해 살펴본다. 특히 unstructured sparsity, structured sparsity, 그리고 최근 제안된 Hierarchical Structured Sparsity (HSS)를 적용한 가속기들을 분석하며, 각 접근 방식의 장단점을 비교한다.

• Journal of Internet Computing and Services
• /
• v.25 no.4
• /
• pp.109-119
• /
• 2024

This study aims to contribute to the initial research on on-device AI in an environment where generative AI-based services on mobile and other on-device platforms are increasing. To derive success strategies for generative AI-based chatbot services in a mobile environment, over 200,000 actual user experience review data collected from the Google Play Store were analyzed using the LDA topic modeling technique. Interpreting the derived topics based on the Information System Success Model (ISSM), the topics such as tutoring, limitation of response, and hallucination and outdated informaiton were linked to information quality; multimodal service, quality of response, and issues of device interoperability were linked to system quality; inter-device compatibility, utility of the service, quality of premium services, and challenges in account were linked to service quality; and finally, creative collaboration was linked to net benefits. Humanization of generative AI emerged as a new experience factor not explained by the existing model. By explaining specific positive and negative experience dimensions from the user's perspective based on theory, this study suggests directions for future related research and provides strategic insights for companies to improve and supplement their services for successful business operations.

• Construction Engineering and Management
• /
• v.25 no.5
• /
• pp.90-92
• /
• 2024

• The monthly packaging world
• /
• s.376
• /
• pp.52-55
• /
• 2024

• Journal of Intelligence and Information Systems
• /
• v.29 no.2
• /
• pp.241-265
• /
• 2023

A corporate insolvency prediction model serves as a vital tool for objectively monitoring the financial condition of companies. It enables timely warnings, facilitates responsive actions, and supports the formulation of effective management strategies to mitigate bankruptcy risks and enhance performance. Investors and financial institutions utilize default prediction models to minimize financial losses. As the interest in utilizing artificial intelligence (AI) technology for corporate insolvency prediction grows, extensive research has been conducted in this domain. However, there is an increasing demand for explainable AI models in corporate insolvency prediction, emphasizing interpretability and reliability. The SHAP (SHapley Additive exPlanations) technique has gained significant popularity and has demonstrated strong performance in various applications. Nonetheless, it has limitations such as computational cost, processing time, and scalability concerns based on the number of variables. This study introduces a novel approach to variable selection that reduces the number of variables by averaging SHAP values from bootstrapped data subsets instead of using the entire dataset. This technique aims to improve computational efficiency while maintaining excellent predictive performance. To obtain classification results, we aim to train random forest, XGBoost, and C5.0 models using carefully selected variables with high interpretability. The classification accuracy of the ensemble model, generated through soft voting as the goal of high-performance model design, is compared with the individual models. The study leverages data from 1,698 Korean light industrial companies and employs bootstrapping to create distinct data groups. Logistic Regression is employed to calculate SHAP values for each data group, and their averages are computed to derive the final SHAP values. The proposed model enhances interpretability and aims to achieve superior predictive performance.

• Annual Conference on Human and Language Technology
• /
• 2021.10a
• /
• pp.482-485
• /
• 2021

최근 지식 그래프를 확장하기 위해 많은 연구가 진행되고 있다. 지식 그래프를 확장하기 위해서는 relation을 기준으로 entity의 방향성을 고려하는 것이 매우 중요하다. 지식 그래프를 확장하기 위한 대표적인 연구인 관계 추출은 문장과 2개의 entity가 주어졌을 때 relation을 예측한다. 최근 사전학습 언어모델을 적용하여 관계 추출에서 높은 성능을 보이고 있지만, entity에 대한 방향성을 고려하여 relation을 예측하는지 알 수 없다. 본 논문에서는 관계 추출에서 entity의 방향성을 고려하여 relation을 예측하는지 실험하기 위해 문장 수준의 Adversarial Attack과 단어 수준의 Sequence Labeling을 적용하였다. 또한 관계 추출에서 문장에 대한 이해를 높이기 위해 BERT모델을 적용하여 실험을 진행하였다. 실험 결과 관계 추출에서 entity에 대한 방향성을 고려하지 않음을 확인하였다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.559-561
• /
• 2022

In order to secure the stability of tap water production and supply, we have built a system that can be pre-verified before applying AI-based composite sensors to the water purification plant, which is a demonstration site. We have collected and analyzed data related to the drug input of the GO-RYEONG water purification plant for about two years from December 2019 to December 2021. The outliers of each tag were removed through data preprocessing such as outliers and derived variable, and the cycle was set as average data for 60 minutes of each one-minute period, and the model was learned using the PLS model.

• Industry Promotion Research
• /
• v.9 no.4
• /
• pp.221-226
• /
• 2024

This study investigates ways to enhance the efficiency of design work in Figma through the use of generative AI. By applying Stephen Anderson's Creating Pleasurable Interface Model, the analysis focuses on six key elements: functional, reliable, usable, convenience, pleasure, and meaningful. In-depth interviews and survey results indicate that Figma's generative AI plugins received generally positive evaluations, particularly for their convenience and usability. However, difficulties in prompt creation and the inconvenience of plugin searches were identified as areas needing improvement. This study provides directions for improving Figma's generative AI capabilities and suggests strategies to enhance the efficiency of design work in practical applications. The study outlines how generative AI can boost designers' creativity and productivity, offering personalized features. These findings serve as a foundation for future design research and practical applications.

• Journal of the Institute of Convergence Signal Processing
• /
• v.23 no.4
• /
• pp.234-240
• /
• 2022

This paper proposes a customized AI exercise recommendation service for balancing the relative amount of exercise according to the working environment by each occupation. WISDM database is collected by using acceleration and gyro sensors, and is a dataset that classifies physical activities into 18 categories. Our system recommends a adaptive exercise using the analyzed activity type after classifying 18 physical activities into 3 physical activities types such as whole body, upper body and lower body. 1 Dimensional convolutional neural network is used for classifying a physical activity in this paper. Proposed model is composed of a convolution blocks in which 1D convolution layers with a various sized kernel are connected in parallel. Convolution blocks can extract a detailed local features of input pattern effectively that can be extracted from deep neural network models, as applying multi 1D convolution layers to input pattern. To evaluate performance of the proposed neural network model, as a result of comparing the previous recurrent neural network, our method showed a remarkable 98.4% accuracy.

• Journal of Venture Innovation
• /
• v.7 no.2
• /
• pp.143-152
• /
• 2024

Recently, as a lot of software with AI functions has been developed, the number of software products with various prediction functions is increasing, and as a result, the importance of software quality has increased. In particular, as consideration for functional safety of products with AI functions increases, software quality management is being conducted at a national level. In particular, the GS Quality Certification System is a quality certification system for software products that is being implemented at the national level, and the GS Certification System is also researching quality evaluation methods for AI products. In this study, we attempt to present an evaluation model that satisfies the basic conditions of software quality based on international standards among the various quality evaluation models presented to verify software reliability. Considering the software quality characteristics of the artificial intelligence sector, we study quality evaluation models, diagnose quality, and predict failures. .In this study, we propose an international standard model for artificial intelligence based on the software reliability growth model, present an evaluation model, and present a method for quality diagnosis through the model. In this respect, this study is considered to be important in that it can predict failures in advance and find failures in advance to prevent risks by predicting the failure time that will occur in software in the future. In particular, it is believed that predicting failures will be important in various safety-related software.