• 한국군사과학기술학회지
• /
• 제21권6호
• /
• pp.865-876
• /
• 2018

This study focuses on the operating requirements of multi-resolution modeling(MRM) in training war-game model and proposes solutions for major issues of multi-resolution interoperation between Combat21 model and tank multi-purpose simulator(TMPS). We study the operating requirements of MRM through interviews with defense M&S experts and literature surveys and report the various issues that could occur with low-resolution model Combat21 and high-resolution model TMPS linked, for example, when to switch objects, what information to exchange, what format to switch to, and how to match data resolutions. This study also addresses the purpose and concept of training using multi-resolution interoperation, role of each model included in multi-resolution interoperation, and issue of matching damage assessments when interoperated between models with different resolutions. This study will provide the common goals and directions of MRM research to MRM researchers, defense modeling & simulation organizations and practitioners.

• 방송공학회논문지
• /
• 제27권2호
• /
• pp.174-184
• /
• 2022

3차원 휴먼 자세 추정 모델은 다시점 모델과 단시점 모델로 분류될 수 있다. 일반적으로 다시점 모델은 단시점 모델에 비하여 뛰어난 자세 추정 성능을 보인다. 단시점 모델의 경우 3차원 자세 추정 성능의 향상은 많은 양의 학습 데이터를 필요로 한다. 하지만 3차원 자세에 대한 참값을 획득하는 것은 쉬운 일이 아니다. 이러한 문제를 다루기 위해, 우리는 다시점 모델로부터 다시점 휴먼 자세 데이터에 대한 의사 참값을 생성하고, 이를 단시점 모델의 학습에 활용하는 방법을 제안한다. 또한, 우리는 각각의 다시점 영상으로부터 추정된 자세의 일관성을 고려하는 다시점 일관성 손실함수를 제안하여, 이것이 단시점 모델의 효과적인 학습에 도움을 준다는 것을 보인다. Human3.6M과 MPI-INF-3DHP 데이터셋을 사용한 실험은 제안하는 방법이 3차원 휴먼 자세 추정을 위한 단시점 모델의 학습에 효과적임을 보여준다.

• Journal of Information Processing Systems
• /
• 제17권3호
• /
• pp.441-452
• /
• 2021

Federated learning provides an efficient integrated model for distributed data, allowing the local training of different data. Meanwhile, the goal of multi-task learning is to simultaneously establish models for multiple related tasks, and to obtain the underlying main structure. However, traditional federated multi-task learning models not only have strict requirements for the data distribution, but also demand large amounts of calculation and have slow convergence, which hindered their promotion in many fields. In our work, we apply the rank constraint on weight vectors of the multi-task learning model to adaptively adjust the task's similarity learning, according to the distribution of federal node data. The proposed model has a general framework for solving optimal solutions, which can be used to deal with various data types. Experiments show that our model has achieved the best results in different dataset. Notably, our model can still obtain stable results in datasets with large distribution differences. In addition, compared with traditional federated multi-task learning models, our algorithm is able to converge on a local optimal solution within limited training iterations.

• The Journal of the Acoustical Society of Korea
• /
• 제29권2E호
• /
• pp.100-106
• /
• 2010

Multiple-model speech recognizer has been shown to be quite successful in noisy speech recognition. However, its performance has usually been tested using the general speech front-ends which do not incorporate any noise adaptive algorithms. For the accurate evaluation of the effectiveness of the multiple-model frame in noisy speech recognition, we used the state-of-the-art front-ends and compared its performance with the well-known multi-style training method. In addition, we improved the multiple-model speech recognizer by employing N-best reference HMMs for interpolation and using multiple SNR levels for training each of the reference HMM.

• 음성과학
• /
• 제15권2호
• /
• pp.55-65
• /
• 2008

Recently, the multi-model based speech recognizer has been used quite successfully for noisy speech recognition. For the selection of the reference HMM (hidden Markov model) which best matches the noise type and SNR (signal to noise ratio) of the input testing speech, the estimation of the SNR value using the VAD (voice activity detection) algorithm and the classification of the noise type based on the GMM (Gaussian mixture model) have been done separately in the multi-model framework. As the SNR estimation process is vulnerable to errors, we propose an efficient method which can classify simultaneously the SNR values and noise types. The KL (Kullback-Leibler) distance between the single Gaussian distributions for the noise signal during the training and testing is utilized for the classification. The recognition experiments have been done on the Aurora 2 database showing the usefulness of the model compensation method in the multi-model based speech recognizer. We could also see that further performance improvement was achievable by combining the probability density function of the MCT (multi-condition training) with that of the reference HMM compensated by the D-JA (data-driven Jacobian adaptation) in the multi-model based speech recognizer.

• 한국컴퓨터정보학회논문지
• /
• 제24권11호
• /
• pp.51-59
• /
• 2019

웹에서 정보 접근에 대한 폭발적인 주문으로 웹 사용자의 다음 접근 페이지를 예측하는 필요성이 대두되었다. 웹 접근 예측을 위해 마코브(markov) 모델, 딥 신경망, 벡터 머신, 퍼지 추론 모델 등 많은 모델이 제안되었다. 신경망 모델에 기반한 딥러닝 기법에서 대규모 웹 사용 데이터에 대한 학습 시간이 엄청 길어진다. 이 문제를 해결하기 위하여 딥 신경망 모델에서는 학습을 여러 컴퓨터에 동시에, 즉 병렬로 학습시킨다. 본 논문에서는 먼저 스파크 클러스터에서 다층 Perceptron 모델을 학습 시킬 때 중요한 데이터 분할, shuffling, 압축, locality와 관련된 기본 파라미터들이 얼마만큼 영향을 미치는지 살펴보았다. 그 다음 웹 접근 예측을 위해 다층 Perceptron 모델을 학습 시킬 때 성능을 높이기 위하여 이들 스파크 파라미터들을 튜닝 하였다. 실험을 통하여 논문에서 제안한 스파크 파라미터 튜닝을 통한 웹 접근 예측 모델이 파라미터 튜닝을 하지 않았을 경우와 비교하여 웹 접근 예측에 대한 정확성과 성능 향상의 효과를 보였다.

• 대기
• /
• 제17권4호
• /
• pp.327-337
• /
• 2007

In this study, we examined the new ensemble training approach to reduce the systematic error and improve prediction skill of wind by using the Short-range Ensemble prediction system (SENSE), which is the mesoscale multi-model ensemble prediction system. The SENSE has 16 ensemble members based on the MM5, WRF ARW, and WRF NMM. We evaluated the skill of surface wind prediction compared with AWS (Automatic Weather Station) observation during the summer season (June - August, 2006). At first stage, the correction of initial state for each member was performed with respect to the observed values, and the corrected members get the training stage to find out an adaptive weight function, which is formulated by Root Mean Square Vector Error (RMSVE). It was found that the optimal training period was 1-day through the experiments of sensitivity to the training interval. We obtained the weighted ensemble average which reveals smaller errors of the spatial and temporal pattern of wind speed than those of the simple ensemble average.

• 한국물환경학회지
• /
• 제27권1호
• /
• pp.30-35
• /
• 2011

2D pollutant transport model was applied to the simulation of contaminant transport in the channel. At first, two kinds of virtual channels having different slopes were designed. The distribution of contaminant, which flows from one of the three drainages to the main channel, was simulated by each 2D model. Concentrations of 745 nodes were converted to input data of neural network model (Multi-perceptron) for training and verification using matrix. The first three cases (Case A-1, A-2, A-3) were used for training Multi-perceptron, the other three cases (Case B-1, B-2, B-3) were used for verification. As a result, Multi-perceptron reasonably divided the cases into the three characteristics which have different contaminant distributions due to the different input point of water pollution source. It can be a useful methodology for the water quality monitoring and backtracking.

• Journal of Electrical Engineering and Technology
• /
• 제11권2호
• /
• pp.265-273
• /
• 2016

A water wall system is one of the most important components of a boiler in a thermal power plant, and it is a nonlinear Multi-Input and Multi-Output (MIMO) system, with 6 inputs and 3 outputs. Three models are developed and comp for the controller design, including a linear model, a multilayer feed-forward neural network (MFNN) model and an Echo State Network (ESN) model. First, the linear model is developed by linearizing a given nonlinear model and is analyzed as a function of the operating point. Second, the MFNN and the ESN are developed by using training data from the nonlinear model. The three models are validated using Matlab with nonlinear input-output data that was not used during training.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권10호
• /
• pp.3668-3684
• /
• 2021

Video action recognition is widely used in video surveillance, behavior detection, human-computer interaction, medically assisted diagnosis and motion analysis. However, video action recognition can be disturbed by many factors, such as background, illumination and so on. Two-stream convolutional neural network uses the video spatial and temporal models to train separately, and performs fusion at the output end. The multi segment Two-Stream convolutional neural network model trains temporal and spatial information from the video to extract their feature and fuse them, then determine the category of video action. Google Xception model and the transfer learning is adopted in this paper, and the Xception model which trained on ImageNet is used as the initial weight. It greatly overcomes the problem of model underfitting caused by insufficient video behavior dataset, and it can effectively reduce the influence of various factors in the video. This way also greatly improves the accuracy and reduces the training time. What's more, to make up for the shortage of dataset, the kinetics400 dataset was used for pre-training, which greatly improved the accuracy of the model. In this applied research, through continuous efforts, the expected goal is basically achieved, and according to the study and research, the design of the original dual-flow model is improved.