• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.446-448
• /
• 2021

In training a deep learning model, it is crucial to tune various hyperparameters and gain speed and accuracy. While hyperparameters that mathematically induce convergence impact training speed, system parameters that affect host-to-device transfer are also crucial. Therefore, it is important to properly tune and select parameters that influence the data loader as a system parameter in overall time acceleration. We propose an automated framework called Num Worker Tuner (NWT) to address this problem. This method finds the appropriate number of multi-processing subprocesses through the search space and accelerates the learning through the number of subprocesses. Furthermore, this method allows memory efficiency and speed-up by tuning the system-dependent parameter, the number of multi-process spawns.

• IEIE Transactions on Smart Processing and Computing
• /
• v.6 no.1
• /
• pp.21-26
• /
• 2017

In this paper, we suggest an automated sleep scoring method using machine learning algorithms on accelerometer data from a wristband device. For an experiment, 36 subjects slept for about eight hours while polysomnography (PSG) data and accelerometer data were simultaneously recorded. After the experiments, the recorded signals from the subjects were preprocessed, and significant features for sleep stages were extracted. The extracted features were classified into each sleep stage using five machine learning algorithms. For validation of our approach, the obtained results were compared with PSG scoring results evaluated by sleep clinicians. Both accuracy and specificity yielded over 90 percent, and sensitivity was between 50 and 80 percent. In order to investigate the relevance between features and PSG scoring results, information gains were calculated. As a result, the features that had the lowest and highest information gain were skewness and band energy, respectively. In conclusion, the sleep stages were classified using the top 10 significant features with high information gain.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.24 no.6A
• /
• pp.775-781
• /
• 1999

A multi-media variable processing gain DS/CDMA system are considered. Two types of information sources with different rates and transmitting powers are assumed to be stransmitted simultaneously in the same channel. Average signal-to-noise ratios at the correlation receiver outputs for each type of information sources are analytically derived as funtions of partial cross-correlations between spreading code sequences. As the difference of information rates between information sources increase, the difference between signal-to-noise ratio regarding random spreading code and that regarding pseudo random spreading code increases (maximum 1.3dB). The result can provide a analytical tools for use in multi-media DS/CDMA system design.

• ETRI Journal
• /
• v.20 no.2
• /
• pp.133-148
• /
• 1998

The objective of this paper is to present an adaptive algorithm for computing the weight vector which provides a beam pattern having its maximum gain along the direction of the mobile target signal source in the presence of interfering signals within a cell. The conjugate gradient method (CGM) is modified in such a way that the suboptimal weight vector is produced with the computational load of O(16N), which has been found to be small enough for the real-time processing of signals in most land mobile communications with the digital signal processor (DSP) off the shelf, where N denotes the number of antenna elements of the array. The adaptive procedure proposed in this paper is applied to code division multiple access (CDMA) mobile communication system to show its excellent performance in terms of signal to interference plus noise ratio (SINR), bit error rate (BER), and capacity, which are enhanced by about 7 dB, ${\frac{1}{100}}$ times, and 7 times, respectively, when the number of antenna elements is 6 and the processing gain is 20 dB.

• Journal of the Institute of Convergence Signal Processing
• /
• v.8 no.3
• /
• pp.217-221
• /
• 2007

This paper deals with the robust nonlinear controller design using output feedback for a Chua circuit which is one of the well-known nonlinear models. First, an exosystem for reference signal tracking is defined, and error dynamic equations are derived from the differentiation of the output tracking error equation. The normal sliding surface is modified using the integral type servo compensator. The parameters in the equations of the modified sliding surface and servo compensator are determined by using the Hurwitz condition of stability. Especially the error signals can't be obtained directly from the output because all parameters are assumed unknown. So instead, a high gain observer is designed. From this estimated error signals, a stabilizing controller is designed. Simulation is done for demonstrating the effectiveness of the suggested algorithm.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.33 no.7
• /
• pp.59-69
• /
• 2005

A position domain Hatch filter is proposed as an efficient carrier-smoothed-code processing algorithm for real-time kinematic differential global satellite navigation systems. The well-known range domain Hatch filter is newly interpreted with a stochastical point of view. The interpretation result is extended to derive the position domain Hatch filter. By a covariance simulation, it is shown that Hatch gain is, in general, more efficient than Kalman-type gain in carrier-smoothed-code processing and the proposed position domain Hatch filter is more advantageous than the conventional range domain Hatch filter if the visible satellite constellation changes during the positioning task.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.9C
• /
• pp.1219-1226
• /
• 2004

The objective of this paper is to proposed a new Hybrid Interference Cancellation(HlC) receiver to cancel MAI in a multirate DS-CDMA system based on multiple processing gain(MPG). We propose a new improved HIC scheme that divides the active users with different data rates split into a number of groups for effectives cancellation Between each group, GW-PIC is performed to cancel other group signals and within them, SIC is carried out to remove multiple access interference in group. We analyze the performance of the proposed receiver in terms of the bit error rate(BER) and examine its performance. As a conclusion, computer simulations show that the proposed schemes outperforms adaptive multistage PIC and conventional SIC receiver over AWGN channel.

• Journal of the Institute of Convergence Signal Processing
• /
• v.5 no.4
• /
• pp.301-307
• /
• 2004

This paper suggests a fuzzy variable structure control scheme for Takagi-Sugeno(T-S) fuzzy model and presents the attitude control of the wheel-driven inverted pendulum(WDIP) based on the proposed control algorithm. The proposed controller is designed based on the T-S fuzzy modeling of nonlinear system and the unification of gain matrices in linear subsystems that constitute the overall fuzzy model. The uncertainties generated in the gain matrix unifying procedure can be interpreted as the input disturbances of the conventional variable structure control. These unifying disturbances can be resolved by using the robustness property of the conventional variable structure system. Design example for wheel-driven inverted pendulum demonstrates the utility and validity of the proposed control scheme.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.05a
• /
• pp.267-269
• /
• 2019

There are many recommendation systems offer an effort to get better preciseness the information to the users. In order to further improve more accuracy, the social network analysis method which is used to analyze data to community detection in social networks was introduced in the recommendation system and the result shows this method is improving more accuracy. In this paper, we propose a movie recommendation system using social network analysis and normalized discounted cumulative gain with the best accuracy. To estimate the performance, the collaborative filtering using the k nearest neighbor method, the social network analysis with collaborative filtering method and the proposed method are used to evaluate the MovieLens data. The performance outputs show that the proposed method get better the accuracy of the movie recommendation system than any other methods used in this experiment.

• Korean Journal of Optics and Photonics
• /
• v.16 no.6
• /
• pp.527-534
• /
• 2005

We investigate the pulsation characteristics in a multi-section DFB laser which is composed of one DFB section, phase tuning section, and gain section. Multi-section DFB lasers with anti-phase (AP) complex-coupled (CC) DFB structure show wide current ranges of gain and phase tuning sections fer stable pulsations compared to those with in-phase CC DFB structure or index-coupled DFB structure. For multi-section DFB lasers with AP CC DFB structure, the current range of a gain section for stable pulsations increases and the tuning range of the pulsation frequency increases as a coupling strength or a gain coupling coefficient increases Also, the tuning range using the phase variation in a phase tuning section increases. For a fixed coupling strength, the current ranges of gain and phase tuning sections for stable pulsations increase and the tuning range of the pulsation frequency increases as the length of a DFB section increases.