• Investigative Magnetic Resonance Imaging
• /
• 제25권1호
• /
• pp.10-22
• /
• 2021

Purpose: To overcome the difficulty in building a large data set with a high-quality in medical imaging, a concept of 'blended-transfer learning' (BTL) using a combination of both source data and target data is proposed for the target task. Materials and Methods: Source and target tasks were defined as training of the source and target networks to reconstruct cardiac CINE images from undersampled data, respectively. In transfer learning (TL), the entire neural network (NN) or some parts of the NN after conducting a source task using an open data set was adopted in the target network as the initial network to improve the learning speed and the performance of the target task. Using BTL, an NN effectively learned the target data while preserving knowledge from the source data to the maximum extent possible. The ratio of the source data to the target data was reduced stepwise from 1 in the initial stage to 0 in the final stage. Results: NN that performed BTL showed an improved performance compared to those that performed TL or standalone learning (SL). Generalization of NN was also better achieved. The learning curve was evaluated using normalized mean square error (NMSE) of reconstructed images for both target data and source data. BTL reduced the learning time by 1.25 to 100 times and provided better image quality. Its NMSE was 3% to 8% lower than with SL. Conclusion: The NN that performed the proposed BTL showed the best performance in terms of learning speed and learning curve. It also showed the highest reconstructed-image quality with the lowest NMSE for the test data set. Thus, BTL is an effective way of learning for NNs in the medical-imaging domain where both quality and quantity of data are always limited.

• Computers and Concrete
• /
• 제32권6호
• /
• pp.625-637
• /
• 2023

The application of ML approaches in determining the resisting capacity of fire damaged RC columns is introduced in this paper, on the basis of analysis data driven ML modeling. Considering the characteristics of the structural behavior of fire damaged RC columns, the representative five approaches of Kernel SVM, ANN, RF, XGB and LGBM are adopted and applied. Additional partial monotonic constraints are adopted in modelling, to ensure the monotone decrease of resisting capacity in RC column with fire exposure time. Furthermore, additional suggestions are also added to mitigate the heterogeneous composition of the training data. Since the use of ML approaches will significantly reduce the computation time in determining the resisting capacity of fire damaged RC columns, which requires many complex solution procedures from the heat transfer analysis to the rigorous nonlinear analyses and their repetition with time, the introduced ML approach can more effectively be used in large complex structures with many RC members. Because of the very small amount of experimental data, the training data are analytically determined from a heat transfer analysis and a subsequent nonlinear finite element (FE) analysis, and their accuracy was previously verified through a correlation study between the numerical results and experimental data. The results obtained from the application of ML approaches show that the resisting capacity of fire damaged RC columns can effectively be predicted by ML approaches.

• Journal of Biosystems Engineering
• /
• 제37권3호
• /
• pp.192-200
• /
• 2012

Purpose: Aerial application of chemicals with an agricultural helicopter allows for precise and timely spraying and reduces working labor and pollution. An attitude controller for an agricultural helicopter would be helpful to aerial application operator. The objectives of this paper are to determine the transfer function models and to estimate the handling qualities of a bare-airframe model helicopter. Methods: Transfer functions of a model unmanned helicopter were estimated by using NAVFIT and DERIVID modules of the $CIFER^{(R)}$ program to the time history data of frequency sweep flight tests. Control inputs of the transfer functions were elevator, aileron, rudder and collective pitch stick positions and the outputs were resulting on-axis movements of the fuselage. Results: Minimum realization of the transfer functions for pitch rate output to elevator control input and roll rate output to aileron control input produced second order transfer functions with undamped natural frequencies around 3.0 Hz and damping ratios of 0.139 and 0.530, respectively. The equivalent time delays of the transfer functions ranged from 0.16 to 0.44 second. Sensitivity analysis of the proposed parameters allowed derivation of minimal realization of the transfer functions. Conclusions: Handling quality of the model helicopter was addressed based on the eigenvalues of the transfer functions, corresponding undamped natural frequencies with damping ratios. The equivalent time delays of the lateral-directional motion ranged from 0.16 to 0.44 second, longer than the 0.1 to 0.15 second requirement for well-controlled typical manned aerial vehicles.

• 정보과학회 논문지
• /
• 제42권12호
• /
• pp.1495-1502
• /
• 2015

센서 네트워크에서 압축은 종단 간 지연시간과 에너지 사용량 사이의 이율배반적인 관계가 있다. 데이터의 크기를 줄이기 위해 압축을 하면, 추가적인 지연시간과 에너지 소비가 발생하지만, 전송으로 인한 에너지 소모는 감소하게 된다. 일반적으로, 배터리기반 센서 네트워크에서는 지연시간을 손해 보더라도 네트워크의 생존시간을 최대화하기위해 압축을 널리 사용하고 있다. 한편 태양 에너지 기반 센서 네트워크에서는 주기적으로 에너지 재생산이 이루어짐에 따라, 동작하는데 충분한 에너지양 이상의 에너지가 존재할 가능성이 있다. 본 논문에서는 여분의 에너지를 사용해 종단 간 지연시간을 감소시키는 에너지 적응형 선택적 압축 기법을 제안한다. 제안하는 기법은 노드의 에너지가 충분하지 않을 때, 에너지 소비를 줄이기 위해 압축을 사용하고, 에너지가 충분한 경우에는 종단 간 지연시간 감소를 위해 압축을 사용하지 않는다. 시뮬레이션을 통한 에너지와 지연시간의 성능평가를 통하여 제안하는 기법의 우수성을 증명하였다.

• 대한용접접합학회:학술대회논문집
• /
• 대한용접접합학회 2009년 추계학술발표대회
• /
• pp.48-48
• /
• 2009

Conventionally in pulsed gas metal arc welding (GMAW-P), drop transfer is analyzed with simplest square pulse waveform. While the pulse current is described by four parameters (peak current magnitude and time plus base current magnitude and time), it deviates the real pulse shape. Real pulse can be better idealized by the trapezoidal pulse waveform described by two additional parameters, i.e., current rise and fall rate (dI/dt). Power source response rate is described by these parameters. In this work, the effect of these parameters on drop transfer is predicted by the force displacement model (FDM). While peak current has significant effects on drop detachment, drop transfer is also influenced by the current rise rate. Predictions indicate that the current rise rate can have considerable effects on the size of the detached drop if other pulse parameters are kept constant. FDM is applied to determine peak time for one drop one pulse condition (ODOP) when rests of the pulse parameters are given. The predicted range of ODOP shows good agreement with experimental data.

• 한국응급구조학회지
• /
• 제27권2호
• /
• pp.51-63
• /
• 2023

Purpose: The study aimed to analyze the occurrence and transfer status of emergency patients at Incheon International Airport in Korea. Methods: This study design included 810 patients for eight years, from July 1, 2014 to June 31, 2022. The collected data were analyzed with SPSS statistics version 25.0. Results: For the demographics, 531 (65.6%) were males, 151 (19.5%) were in the highest age group (in their 50s), and 289 (35.9%) were foreigners. The most common place of occurrence was in protected areas in the passenger terminal, with 341 (42.1%) of cases occurring here. The time of arrival at the site (z=-3.444, p=.001), stay duration at the site (z=-8.145, p=. 001), and transfer time (z=-3.623, p=.000) were all significantly longer. Conclusion: It is necessary to consider a rapid emergency transportation plan, such as developing a system that simplifies immigration procedures when emergency patients move from protected areas to general areas. Moreover, in order to quickly respond to foreign emergency patients, detailed characteristics will need to be identified. Lastly, considering the long transfer time, an instructing doctor should accompany an ambulance and actively perform advanced resuscitation.

• Journal of Information Processing Systems
• /
• 제19권5호
• /
• pp.673-687
• /
• 2023

Convolutional neural network-based deep learning technology is the most commonly used in image identification, but it requires large-scale data for training. Therefore, application in specific fields in which data acquisition is limited, such as in the military, may be challenging. In particular, the identification of ground weapon systems is a very important mission, and high identification accuracy is required. Accordingly, various studies have been conducted to achieve high performance using small-scale data. Among them, the ensemble method, which achieves excellent performance through the prediction average of the pre-trained models, is the most representative method; however, it requires considerable time and effort to find the optimal combination of ensemble models. In addition, there is a performance limitation in the prediction results obtained by using an ensemble method. Furthermore, it is difficult to obtain the ensemble effect using models with imbalanced classification accuracies. In this paper, we propose a transfer learning-based feature fusion technique for heterogeneous models that extracts and fuses features of pre-trained heterogeneous models and finally, fine-tunes hyperparameters of the fully connected layer to improve the classification accuracy. The experimental results of this study indicate that it is possible to overcome the limitations of the existing ensemble methods by improving the classification accuracy through feature fusion between heterogeneous models based on transfer learning.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.1260-1263
• /
• 2005

Data storage related with writing and retrieving requires high storage capacity, fast transfer rate and less access time. Today any data storage system cannot satisfy these conditions, however holographic data storage system can perform faster data transfer rate because it is a page oriented memory system using volume hologram in writing and retrieving data. System can be constructed without mechanical actuating part therefore fast data transfer rate and high storage capacity about 1Tb/cm3 can be realized. In this research, to reduce errors of binary data stored in holographic data storage system, a new method for bit error reduction is suggested. First, find fuzzy rule using experimental system for Element of Holographic Digital Data System. Second, make fuzzy rule table using Genetic algorithm. Third, reduce prior error element and recording Digital Data. Recording ratio and reconstruction ratio will be very good performance

• 정보저장시스템학회:학술대회논문집
• /
• 정보저장시스템학회 2005년도 추계학술대회 논문집
• /
• pp.44-46
• /
• 2005

Data storage related with writing and retrieving requires high storage capacity, fast transfer rate and less access time. Today any data storage system cannot satisfy these conditions, however holographic data storage system can perform faster data transfer rate because it is a page oriented memory system using volume hologram in writing and retrieving data. System can be constructed without mechanical actuating part therefore fast data transfer rate and high storage capacity about $1Tb/cm^3$ can be realized. In this paper, to reduce errors of binary data stored in holographic data storage system, a new method for bit error reduction is suggested. We proposal Algorithm use The Extended Kalman filter. The Kalman filter reduce measurement noise. Therefore, By using this error reduction method following results are obtained; the effect of measurement nois of Pixel is decreased and the intensity profile of data page becomes uniform therefore the better data storage system can be constructed.

• 정보저장시스템학회:학술대회논문집
• /
• 정보저장시스템학회 2005년도 추계학술대회 논문집
• /
• pp.129-133
• /
• 2005

Data storage related with writing and retrieving requires high storage capacity, fast transfer rate and less access time. Today any data storage system cannot satisfy these conditions, however holographic data storage system can perform faster data transfer rate because it is a page oriented memory system using volume hologram in writing and retrieving data. System can be constructed without mechanical actuating part therefore fast data transfer rate and high storage capacity about $1Tb/cm^3$ can be realized. In this paper, to reduce errors of binary data stored in holographic data storage system, a new method for bit error reduction is suggested. First, find cluster centers using subtractive clustering algorithm then reduce intensities of pixels around cluster centers and fuzzy rules. Therefore, By using this error reduction method following results are obtained ; the effect of Inter Pixel Interference noise is decreased and the intensity profile of data page becomes uniform therefore the better data storage system can be constructed.