163559 research outputs found
Sort by
An organic approach
No full textCopper catalysts hold promise for producing multi-carbon chemicals through electrochemical CO2 reduction, but improving performance is challenging due to the limited tunability of the copper surface. Now, research uses organic functionalization to modify the surface oxidation state of copper, yielding improved energy efficiency for ethylene production.N
2Memristor-1Capacitor Integrated Temporal Kernel for High-Dimensional Data Mapping
No full textCompact but precise feature-extracting ability is core to processing complex computational tasks in neuromorphic hardware. Physical reservoir computing (RC) offers a robust framework to map temporal data into a high-dimensional space using the time dynamics of a material system, such as a volatile memristor. However, conventional physical RC systems have limited dynamics for the given material properties, restricting the methods to increase their dimensionality. This study proposes an integrated temporal kernel composed of a 2-memristor and 1-capacitor (2M1C) using a W/HfO2/TiN memristor and TiN/ZrO2/Al2O3/ZrO2/TiN capacitor to achieve higher dimensionality and tunable dynamics. The kernel elements are carefully designed and fabricated into an integrated array, of which performances are evaluated under diverse conditions. By optimizing the time dynamics of the 2M1C kernel, each memristor simultaneously extracts complementary information from input signals. The MNIST benchmark digit classification task achieves a high accuracy of 94.3% with a (196x10) single-layer network. Analog input mapping ability is tested with a Mackey-Glass time series prediction, and the system records a normalized root mean square error of 0.04 with a 20x1 readout network, the smallest readout network ever used for Mackey-Glass prediction in RC. These performances demonstrate its high potential for efficient temporal data analysis. An integrated temporal kernel using two memristors and one capacitor is fabricated. This kernel extracts complementary features from the input, ultimately processing MNIST images at 8-bit to achieve an accuracy of 94.3%. Excellent prediction performance for the Mackey-Glass time series is verified with NRMSE of 0.04 in minimal network size (20 x 1).imageN
High-Pressure Deuterium Annealing for Trap Passivation for a 3-D Integrated Structure
No full textHigh-pressure deuterium annealing (HPDA) and forming gas annealing (FGA) were applied to monolithically and vertically integrated MOSFETs with a 3-D architecture of one over the other. An overlying poly-Si thin-film transistor (TFT) is positioned over an underlying MOSFET onto a wafer of silicon-on-insulator (SOI). The effects of HPDA and FGA on these double-stacked MOSFETs were quantitatively analyzed by extracting the interface trap density (N-it) from dc I-V characteristics and border trap density (N-bt) through low-frequency noise (LFN) measurements. The performance index parameters, such as subthreshold swing (SS) and on-state current (I-ON), were also comparatively analyzed. It has been confirmed that, for the superjacent MOSFET, HPDA reduced N-it by 250% and N-bt by 92% compared to FGA. Additionally, for the subjacent MOSFET, HPDA decreased N-it by 15% and Nbt by 32% compared to FGA.N
Cross-Wired Memristive Crossbar Array for Effective Graph Data Analysis
No full textGraphs adequately represent the enormous interconnections among numerous entities in big data, incurring high computational costs in analyzing them with conventional hardware. Physical graph representation (PGR) is an approach that replicates the graph within a physical system, allowing for efficient analysis. This study introduces a cross-wired crossbar array (cwCBA), uniquely connecting diagonal and non-diagonal components in a CBA by a cross-wiring process. The cross-wired diagonal cells enable cwCBA to achieve precise PGR and dynamic node state control. For this purpose, a cwCBA is fabricated using Pt/Ta2O5/HfO2/TiN (PTHT) memristor with high on/off and self-rectifying characteristics. The structural and device benefits of PTHT cwCBA for enhanced PGR precision are highlighted, and the practical efficacy is demonstrated for two applications. First, it executes a dynamic path-finding algorithm, identifying the shortest paths in a dynamic graph. PTHT cwCBA shows a more accurate inferred distance and approximate to 1/3800 lower processing complexity than the conventional method. Second, it analyzes the protein-protein interaction (PPI) networks containing self-interacting proteins, which possess intricate characteristics compared to typical graphs. The PPI prediction results exhibit an average of 30.5% and 21.3% improvement in area under the curve and F1-score, respectively, compared to existing algorithms. This study introduces a cross-wired crossbar array (cwCBA) using a Pt/Ta2O5/HfO2/TiN memristor for precise physical graph representation (PGR) and dynamic node state control. The cwCBA, with enhanced PGR precision, successfully executes a dynamic path-finding algorithm with significantly lower processing complexity and analyzes complex protein-protein interaction networks, showing notable improvements in predictive accuracy over conventional methods.imageN
Remote Epitaxy: Fundamentals, Challenges, and Opportunities
No full textAdvanced heterogeneous integration technologies are pivotal for next-generation electronics. Single-crystalline materials are one of the key building blocks for heterogeneous integration, although it is challenging to produce and integrate these materials. Remote epitaxy is recently introduced as a solution for growing single-crystalline thin films that can be exfoliated from host wafers and then transferred onto foreign platforms. This technology has quickly gained attention, as it can be applied to a wide variety of materials and can realize new functionalities and novel application platforms. Nevertheless, remote epitaxy is a delicate process, and thus, successful execution of remote epitaxy is often challenging. Here, we elucidate the mechanisms of remote epitaxy, summarize recent breakthroughs, and discuss the challenges and solutions in the remote epitaxy of various material systems. We also provide a vision for the future of remote epitaxy for studying fundamental materials science, as well as for functional applications.N
PAC-FNO: PARALLEL-STRUCTURED ALL-COMPONENT FOURIER NEURAL OPERATORS FOR RECOGNIZING LOW-QUALITY IMAGES
No full textA standard practice in developing image recognition models is to train a model on a specific image resolution and then deploy it. However, in real-world inference, models often encounter images different from the training sets in resolution and/or subject to natural variations such as weather changes, noise types and compression artifacts. While traditional solutions involve training multiple models for different resolutions or input variations, these methods are computationally expensive and thus do not scale in practice. To this end, we propose a novel neural network model, parallel-structured and all-component Fourier neural operator (PAC-FNO), that addresses the problem. Unlike conventional feed-forward neural networks, PAC-FNO operates in the frequency domain, allowing it to handle images of varying resolutions within a single model. We also propose a two-stage algorithm for training PAC-FNO with a minimal modification to the original, downstream model. Moreover, the proposed PAC-FNO is ready to work with existing image recognition models. Extensively evaluating methods with seven image recognition benchmarks, we show that the proposed PAC-FNO improves the performance of existing baseline models on images with various resolutions by up to 77.1% and various types of natural variations in the images at inference.N
Next-Generation Nitrate, Ammonium, Phosphate, and Potassium Ion Monitoring System in Closed Hydroponics: Review on State-of-the-Art Sensors and Their Applications
No full textClosed hydroponics is an environmentally friendly and economical method for growing crops by circulating a nutrient solution while measuring and supplementing various ions contained in the solution. However, conventional monitoring systems in hydroponics do not measure individual ions in the nutrient solution; instead, they predict the total ion content from the pH and electrical conductivity (EC). This method cannot be used to supplement individual ions and adjusts the concentration of the circulating nutrient solution by diluting or supplying a premixed nutrient solution. A more advanced system should be able to identify the concentration of each ion in the nutrient solution and supplement any deficient ions, thus requiring individual ion monitoring systems. Therefore, we first investigated the nitrate, ammonium, phosphate, and potassium (NPK) ion concentration and pH range commonly used for nutrient solutions. Subsequently, we discuss the latest research trends in electrochemical and optical sensors for measuring NPK ions. We then compare the conventional monitoring system (pH and EC-based) and advanced monitoring systems (individual ion sensors) and discuss the respective research trends. In conclusion, we present the hurdles that researchers must overcome in developing agricultural ion sensors for advanced monitoring systems and propose the minimum specifications for agricultural NPK ion sensors.Y
Microneedle-Based Precision Engineering for Agrochemical Delivery and Plant Health Monitoring <i>In Situ</i>
No full textMicroneedles, which are small needle-shaped devices, have gained attention as versatile tools with applications across various fields owing to their precision and minimal invasiveness. In the literature, the integration of microneedle technology into agricultural domains has been explored, particularly focusing on the real-time monitoring of precise agrochemical delivery and crop biosensing in situ. In this study, we summarize representative microneedle types and corresponding fabrication techniques. Then, we discuss the advantages of microneedles in plant health management and drug delivery, as well as related challenges, such as safety and mass production. The agricultural applications of microneedles can potentially overcome the limitations of traditional methods and offer innovative solutions for crop disease management. This study demonstrates that microneedles present innovative possibilities in agriculture and contribute to the improvement of sustainable agriculture.N
Impact of perioperative high-intensity statin treatment on the occurrence of postoperative atrial fibrillation after coronary artery bypass grafting: ameta-analysis
No full textBackground: This meta-analysis was conducted to evaluate the impact of high-intensity statin treatment on new-onset postoperative atrial fibrillation (POAF) after coronary artery bypass grafting (CABG). Methods: Four databases were searched for studies that enrolled patients who underwent CABG and investigated the impact of perioperative use of high-intensity statins on the occurrence rate of POAF. The primary outcome was the incidence of POAF. Secondary outcomes were operative mortality and perioperative myocardial infarction (PMI). Publication bias was assessed using a funnel plot and Egger's test. Results: Nine articles (eight randomized controlled trials and one non-randomized study: n=3,072) were selected. Rosuvastatin (20 mg) was used in four studies, while atorvastatin (40-80 mg) was used in the other five studies. Reported incidences of POAF in the included studies ranged from 11% to 48.8%. Pooled analyses showed that the incidence of POAF was significantly lower in patients treated with high-intensity statins than in patients in the control group patients (odds ratio, 0.43; 95% CI, 0.27-0.68; P<0.001). Subgroup analyses showed that the impact of high-intensity statins was significant in studies using atorvastatin but not in studies using rosuvastatin. There was no significant subgroup difference in the primary endpoint between studies using a placebo and those using low-dose statins. Secondary outcomes, including operative mortality and the incidence of PMI, were not affected by high-intensity statin treatment. Conclusions: Perioperative use of high-intensity statins is associated with a 57% reduction in the occurrence of POAF among patients undergoing CABG.Y
Development and Evaluation of Nursing Care Needs Machine Learning-based Prediction Model Using Medical Procedure Codes in Comprehensive Nursing Care Wards in a Tertiary Hospital
No full text학위논문(석사) -- 서울대학교 대학원 : 간호대학 간호학과, 2024. 8. 서은영.입원환자에게 질적인 간호를 제공하기 위해서 간호인력을 적정하게 배치하는 것이 필요하다. 적정 간호인력 배치수준을 결정하기 위해서 간호필요도(nursing care needs)를 고려해야한다. 간호필요도란 환자의 건강상태와 관련된 모든 측면에서 간호사의 도움이 필요한 요구사항을 의미한다. 간호필요도 측정도구로 한국형 환자분류도구(Korean Patient Classification System, KPCS)가 여러 의료기관에서 사용되고 있으나 모든 의료기관에서 사용 중인 것은 아니며 KPCS-1을 사용하여 측정한 간호필요도를 축적하는 시스템도 개발되어 있지 않은 실정이다. 전국 모든 의료기관에 입원한 환자의 간호필요도를 추산할 수 있는 모델을 개발한다면 간호인력 배치 수준을 평가하고 이를 바탕으로 효과적인 간호인력 정책을 만들 수 있다. 본 연구는 이러한 모델을 개발하기 위한 기초연구로 일반화 성능이 뛰어난 머신러닝을 활용하여 간호·간병통합서비스 병동의 간호필요도 예측 모델을 개발하였다. 본 연구는 일개 상급종합병원 간호·간병통합서비스 병동에 2019년부터 2022년에 입원한 성인환자 13,828명의 일자별 의료행위 코드와 간호필요도 점수 데이터를 사용하여 XGBoost, Light GBM, Random Forest, Ridge Regression, Lasso Regression 5개의 머신러닝 기반 예측모델을 생성하였다. 모델링은 K-Fold 검증으로 Mean Squared Error, R2 score를 기준으로 모델을 평가하였다. XGBoost의 Train Set 평균 MSE와 R2 점수는 0.3936, 0.9342, Test Set 평균 MSE와 R2 점수는 0.4650, 0.9222로 가장 성능이 뛰어났다. Light GBM의 Train Set 평균 MSE와 R2 점수는 0.4447, 0.9256, Test Set 평균 MSE와 R2 점수는 0.4693, 0.9215이었으며 Random Forest는 Train Set 평균 MSE와 R2 점수는 0.1177, 0.9803, Test Set 평균 MSE와 R2 점수는 0.4981, 0.9167이었다. Ridge Regression의 Train Set 평균 MSE와 R2 점수는 0.5982, 0.9000, Test Set 평균 MSE와 R2 점수는 0.6108, 0.8978이었다. Lasso Regression은 Train Set과 Test Set의 R2 점수가 거의 0점으로 유효하지 않았다. 모델링에 사용한 총 722개의 의료행위 코드 중 간호필요도 점수에 미치는 영향을 확인하기 위하여 순열 중요도를 사용하였다. XGBoost에서 가장 중요한 변수는 Exercise & Activity assist (Fulll), Position change assist (Partial), Exercise & Activity assist (Partial)이, Light GBM은 Exercise & Activity assist (Fulll), Position change with massage (exclude prone position), Position change assist (Partial)이었다. Random Forest는 Exercise & Activity assist (Fulll), Position change with massage (exclude prone position), Exercise & Activity assist (Partial)이 중요한 변수였고 Ridge는 Exercise & Activity assist (Fulll), Position change with massage (exclude prone position), Exercise & Activity assist (Partial), Stoma care (1st stoma)이었다. 모든 모델에서 모델링에 중요한 영향을 미치는 변수는 대부분 ADL 보조와 관련된 행위들이었음을 알 수 있었다. 본 연구는 간호필요도 예측에 머신러닝 알고리즘의 활용 가능성을 확인하고 전국 의료기관에서 공통으로 사용할 수 있는 간호필요도 예측 모델 개발의 기초 연구로 활용될 수 있을 것이다. 추후 연구에서 전국 의료기관에서 공통으로 사용하는 청구 코드를 사용하고 간호·간병통합서비스 병동 뿐만 아니라 일반병동 등으로 대상 병동을 확대한다면 일반화 가능성이 높은 모델을 만들 수 있을 것이다.
주요어 : 간호필요도, 의료행위 코드, 머신러닝, 순열중요도 학 번 : 2022-25845Adequate nursing staffing is necessary to provide quality care to inpatients. To determine the appropriate staffing level, nursing care needs should be considered. Nursing care needs are the requirements for nursing assistance in all aspects of a patient's health condition. The Korean Patient Classification System (KPCS) is used in many nursing institutions to measure nursing care needs, but it is not used in all nursing institutions, and a system for accumulating nursing care needs measured using KPCS-1 has not been developed. Developing a model that can estimate the nursing needs of patients hospitalized in all nursing institutions nationwide can be used to evaluate nursing staffing levels and create effective nursing staffing policies. As a basic study to develop such a model, this study developed a nursing care needs prediction model for nursing-care integrated service wards using machine learning with excellent generalization performance. This study generated five machine learning-based predictive models, XGBoost, Light GBM, Random Forest, Ridge Regression, and Lasso Regression, using date-specific medical procedure codes and nursing need score data of 13,828 adult patients hospitalized in a nursing-care integrated service ward of a tertiary hospital from 2019 to 2022. The models were evaluated based on Mean Squared Error and R2 score using K-Fold validation. XGBoost performed the best with Train Set average MSE and R2 scores of 0.3936 and 0.9342, and Test Set average MSE and R2 scores of 0.4650 and 0.9222. Light GBM had train set mean MSE and R2 scores of 0.4447 and 0.9256, and test set mean MSE and R2 scores of 0.4693 and 0.9215, while Random Forest had train set mean MSE and R2 scores of 0.1177 and 0.9803, and test set mean MSE and R2 scores of 0.4981 and 0.9167. Ridge Regression had Train Set average MSE and R2 scores of 0.5982 and 0.9000, and Test Set average MSE and R2 scores of 0.6108 and 0.8978. Lasso Regression was invalid with an R2 score of almost zero for the Train Set and Test Set. Of the total 722 medical procedure codes used in the modeling, permutation importance was used to determine their impact on the nursing need score. The most important variables for XGBoost were Exercise & Activity assist (Full), Position change assist (Partial), and Exercise & Activity assist (Partial), while for Light GBM they were Exercise & Activity assist (Full), Position change with massage (exclude prone position), and Position change assist (Partial). For Random Forest, Exercise & Activity assist (Full), Position change with massage (exclude prone position), and Exercise & Activity assist (Partial) were the significant variables, and for Ridge, Exercise & Activity assist (Full), Position change with massage (exclude prone position), Exercise & Activity assist (Partial), and Stoma care (1st stoma) were the significant variables. In all models, the variables that had a significant impact on the modeling were mostly activities related to assisting with ADLs. This study confirms the feasibility of using machine learning algorithms to predict nursing care needs and can be used as a basic study for developing a nursing care needs prediction model that can be used in nursing institutions nationwide. Future studies should use billing codes common to nursing homes across the country and expand the target wards to include general wards as well as nursing and integrated care wards to create a model with high generalizability.제 1 장 서 론 9
제 1 절 연구의 필요성 9
제 2 절 연구의 목적 · 11
제 3 절 용어의 정의 · 11
제 2 장 문헌 고찰 14
제 3 장 연구방법 · 20
제 1 절 연구 설계 20
제 2 절 연구 대상 20
제 3 절 연구 도구 20
제 4 절 CRISP-DM에 따른 자료 수집 및 분석 21
제 5 절 연구의 윤리적 고려 27
제 4 장 연구결과 · 29
제 1 절 일반적 특성 · 29
제 2 절 의료행위 코드 31
제 3 절 간호필요도 점수 33
제 4 절 모델링 40
제 5 절 변수 중요도 · 46
제 5 장 논의 · 50
제 1 절 개발한 모델의 성능 50
제 2 절 진료과별 간호필요도 점수 51
제 3 절 간호필요도 점수에 영향을 미치는 의료행위 코드 52
제 4 절 간호필요도 점수 총점 53
제 5 절 입원일, 퇴원일의 간호필요도 점수와 모델의 예측 성능 ·· 53
제 6 절 간호필요도와 간호사 인력 배치 수준 · 54
제 7 절 연구의 제한점 55
제 6 장 결론 및 제언 57
제 1 절 결론 · 57
제 2 절 제언 · 58
참고문헌 60
Abstract 65석