• Journal of Information Processing Systems
• /
• v.17 no.5
• /
• pp.867-878
• /
• 2021

In context awareness and user intention tasks, dataset construction is expensive because specific domain data are required. Although pretraining with a large corpus can effectively resolve the issue of lack of data, it ignores domain knowledge. Herein, we concentrate on data domain knowledge while addressing data scarcity and accordingly propose a multi-channel long short-term memory (LSTM). Because multi-channel LSTM integrates pretrained vectors such as task and general knowledge, it effectively prevents catastrophic forgetting between vectors of task and general knowledge to represent the context as a set of features. To evaluate the proposed model with reference to the baseline model, which is a single-channel LSTM, we performed two tasks: voice phishing with context awareness and movie review sentiment classification. The results verified that multi-channel LSTM outperforms single-channel LSTM in both tasks. We further experimented on different multi-channel LSTMs depending on the domain and data size of general knowledge in the model and confirmed that the effect of multi-channel LSTM integrating the two types of knowledge from downstream task data and raw data to overcome the lack of data.

• The Journal of the Korea Contents Association
• /
• v.9 no.7
• /
• pp.95-101
• /
• 2009

Cognitive Radios are encouraging solutions to improve the utilization of the radio spectrum In this paper, we propose channel allocation model for cognitive radio and analyse the performance of the model with Petri Nets. We design the model with an assumption that only partial information about current channel allocation information is known to CR users. The model is designed to communicate efficiently avoiding interference with primary and CR users and contains channel reobtaining process of CR users. The dropping rate and throughput of CR users under the various channel utilization of primary users are given as performance index.

• JSTS:Journal of Semiconductor Technology and Science
• /
• v.7 no.2
• /
• pp.120-131
• /
• 2007

A comprehensive short channel analytical model has been proposed for High Electron Mobility Transistor (HEMT) to obtain higher cut-off frequency maintaining the reliability of the device. The model has been proposed to consider generalized doping variation in the directions perpendicular to and along the channel. The effect of field plates and different gate-insulator geometry (T-gate, etc) have been considered by dividing the area between gate and the high band gap semiconductor into different regions along the channel having different insulator and metal combinations of different thicknesses and work function with the possibility that metal is in direct contact with the high band gap semiconductor. The variation obtained by gate-insulator geometry and field plates in the field and channel potential can be produced by varying doping concentration, metal work-function and gate-stack structures along the channel. The results so obtained for normal device structure have been compared with previous proposed model and numerical method (finite difference method) to prove the validity of the model.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.45 no.2
• /
• pp.1-6
• /
• 2008

A compact analytical model of the threshold voltage for long-channel Asymmetric Double-Gate(ADG) MOSFET is presented. In contrast to the previous models, channel doping and carrier quantization are taken into account. A more compact model is derived by utilizing the potential distribution linearity characteristic of silicon film at threshold. The accuracy of the model is verified by comparisons with numerical simulations for various silicon film thickness, channel doping concentration and oxide thickness.

• Nuclear Engineering and Technology
• /
• v.54 no.3
• /
• pp.1126-1135
• /
• 2022

The MARS code has been assessed for the prediction of onset of flow instability (OFI) in a vertical channel. For assessment, we built an experiment database that consists of experiments under various geometry and thermal-hydraulic condition. It covers pressure from 0.12 to 1.73 MPa; heat flux from 0.67 to 3.48 MW/m²; inlet sub-cooling from 39 to 166 ℃; hydraulic diameters between 2.37 and 6.45 mm of rectangular channels and pipes. It was shown that the MARS code can predict the OFI mass flux for pipes reasonably well. However, it could not predict the OFI in a rectangular channel well with a mean absolute percentage error of 8.77%. In the cases of rectangular channels, the error tends to depend on the hydraulic diameter. Because the OFI is directly related to the subcooled boiling in a flow channel, we suggest a modified subcooled boiling model for better prediction of OFI in a rectangular channel; the net vapor generation (NVG) model and the modified wall evaporation model were modified so that the effect of hydraulic diameter and heat flux can be accurately considered. The assessment of the modified model shows the prediction of OFI mass flux for rectangular channels is greatly improved.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.42 no.9 s.339
• /
• pp.19-28
• /
• 2005

We establish a statistical model for the ultra-wide bandwidth (UMB) indoor channel based on over 2000 frequency response measurements campaign in a Practical apartment. The approach is based on the investigation of the statistical properties of the multipath profiles measured in different place with different rooms. Based on the experimental results, a characterization of the propagation channel from theoretic view point is described. Also we describe a method for measurement of the channel impulse response and channel transfer function. Using the measured data, the authors compares channel impulse responses obtained from time-domain and channel transfer functions obtained from frequency-domain with statistical path loss model. The bandwidth of the signal used in this experiment is from 10MHz to 8.01 GHz. The time-domain results such as maximum excess delay, men excess delay and ms delay spread are presented. As well as, omni-directional biconical antenna were used for transmitter and receiver In addition, measurements presented here support m channel model including the antenna characteristics.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.15 no.8
• /
• pp.759-764
• /
• 2004

This paper proposes a method that estimates MIMO channel characteristics analytically using a 3D ray tracing propagation model. We calculate the discrete spatial correlation between sub-channels by considering phase differences of paths, and using this, estimate the mean capacity upper bound of MIMO channel by Jensen's inequality. This analysis model is a deterministic model that do not approach stochastically through measurement nor approach statistically through Monte-Carlo simulations, so this model has high efficiency for time and cost. And based on the electromagnetic theory, this model may analyze quantitatively the parameters which can affect the channel capacity - antenna pattern, polarization mutual coupling, antenna structure and etc. This model may be used for the development of an optimal antenna structure for MIMO systems.

• The Journal of Korean Institute of Electromagnetic Engineering and Science
• /
• v.26 no.8
• /
• pp.748-757
• /
• 2015

MIMO systems utilizing multiple antenna transmission and reception is one of the key technologies to enhance the capacity of 5G wireless communications. In order to obtain an appropriate performance evaluation of MIMO techniques, the usage of wireless channel model reflecting spatial channel characteristics is required, such as the 3-dimensional spatial channel model(3D SCM) proposed by 3GPP TR36.873 documentation. In this paper, we implement and verify the channel simulation environment based on 3D SCM, to present and compare the characteristics of UMi and UMa environments. We also apply MIMO transmission to the UMa scenario to investigate the channel correlation among antenna elements with different array distances and to identify the corresponding throughput changes. By evaluating the channel power correlations for randomly distributed users within the sector for different horizontal and vertical antenna distances, we present the statistical characteristics which determine the transmission performance under the SCM environment.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• v.20 no.1
• /
• pp.37-50
• /
• 2016

A large eddy simulation (LES) of a turbulent channel flow is performed by using the third order low-storage Runge-Kutta method in time and second order finite difference formulation in space with staggered grid at a Reynolds number, $Re_{\tau}=590$ based on the channel half width, ${\delta}$ and wall shear velocity, $u_{\tau}$. To reduce the calculation cost of LES, algebraic wall model (AWM) is applied to approximate the near-wall region. The computation is performed in a domain of $2{\pi}{\delta}{\times}2{\delta}{\times}{\pi}{\delta}$ with $32{\times}20{\times}32$ grid points. Standard Smagorinsky model is used for subgrid-scale (SGS) modeling. Essential turbulence statistics of the flow field are computed and compared with Direct Numerical Simulation (DNS) data and LES data using no wall model. Agreements as well as discrepancies are discussed. The flow structures in the computed flow field have also been discussed and compared with LES data using no wall model.

• Korean Journal of Air-Conditioning and Refrigeration Engineering
• /
• v.19 no.12
• /
• pp.831-841
• /
• 2007

This work is to extend the elliptic operator, which has been already adopted in turbulent stress model, to fully developed turbulent buoyant channel flows with changing the orientation of the buoyancy vector to be perpendicular to the channel walls. The turbulent heat flux models based on the elliptic concept are employed and closely linked to the elliptic blending second moment closure which is used for the prediction of Reynolds stresses. In order to reflect the stable or unstable stratification conditions, the present model introduces the gradient Richardson number into the thermal to mechanical time scale ratio and model coefficients. The present model has been applied for the computation of stably and unstably stratified turbulent channel flows and the prediction results are directly compared to the DNS data.