Deep Neural Network 기반 프로야구 일일 관중 수 예측 : 광주-기아 챔피언스 필드를 중심으로

박동주(Dong Ju Park),김병우(Byeong Woo Kim),정영선(Young-Seon Jeong),안창욱(Chang Wook Ahn) 한국스마트미디어학회 2018 스마트미디어저널 Vol.7 No.1

본 연구는 Deep Neural Network(DNN)을 이용하여 광주-기아 챔피언스 필드의 일일 관중 수를 예측함으로써 이를 통해 구단과 관련기업의 마케팅 자료제공 및 구장 내 부대시설의 재고관리에 자료로 쓰임을 목적으로 수행 되었다. 본 연구 에서는 Artificial Neural Network(ANN)의 종류인 DNN 모델을 이용하였으며 DNN 모델의 과적합을 막기 위해 Dropout 과 Batch normalization 적용한 모델을 바탕으로 총 4종류를 설계하였다. 각각 10개의 DNN을 만들어 예측값의 Root Mean Square Error(RMSE)와 Mean Absolute Percentage Error(MAPE)의 평균값을 낸 모델과 예측값의 평균으로 RMSE와 MAPE를 평가한 Ensemble 모델을 만들었다. 모델의 학습 데이터는 2008년부터 2017년까지의 관중 수 데이터를 수집하여 수집된 데이터의 80%를 무작위로 선정하였으며, 나머지 20%는 테스트 데이터로 사용하였다. 총 100회의 데이터 선정, 모델구성 그리고 학습 및 예측을 한 결과 Ensemble 모델은 DNN 모델의 예측력이 가장 우수하게 나왔으며, 다중선 형회귀 모델 대비 RMSE는 15.17%, MAPE는 14.34% 높은 예측력을 보이고 있다. In this paper, we used the Deep Neural Network (DNN) to predict the number of daily spectators of Gwangju -KIA Champions Field in order to provide marketing data for the team and related businesses and for managing the inventories of the facilities in the stadium. In this study, the DNN model, which is based on an artificial neural network (ANN), was used, and four kinds of DNN model were designed along with dropout and batch normalization model to prevent overfitting. Each of four models consists of 10 DNNs, and we added extra models with ensemble model. Each model was evaluated by Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The learning data from the model randomly selected 80% of the collected data from 2008 to 2017, and the other 20% were used as test data. With the result of 100 data selection, model configuration, and learning and prediction, we concluded that the predictive power of the DNN model with ensemble model is the best, and RMSE and MAPE are 15.17% and 14.34% higher, correspondingly, than the prediction value of the multiple linear regression model.

DNN을 이용한 응시 깊이 보정

한석호,장훈석 한국정보전자통신기술학회 2023 한국정보전자통신기술학회논문지 Vol.16 No.3

if we know what we’re looking at, we can get a lot of information. Due to the development of eye tracking, Information on gaze point can be obtained through software provided by various eye tracking equipments. However, it is difficult to estimate accurate information such as the actual gaze depth. If it is possible to calibrate the eye tracker with the actual gaze depth, it will enable the derivation of realistic and accurate results with reliable validity in various fields such as simulation, digital twin, VR, and more. Therefore, in this paper, we experiment with acquiring and calibrating raw gaze depth using an eye tracker and software. The experiment involves designing a Deep Neural Network (DNN) model and then acquiring gaze depth values provided by the software for specified distances from 300mm to 10,000mm. The acquired data is trained through the designed DNN model and calibrated to correspond to the actual gaze depth. In our experiments with the calibrated model, we were able to achieve actual gaze depth values of 297mm, 904mm, 1,485mm, 2,005mm, 3,011mm, 4,021mm, 4,972mm, 6,027mm, 7,026mm, 8,043mm, 9,021mm, and 10,076mm for the specified distances from 300mm to 10,000mm. 응시점을 통해 어떤 것을 보고 있는지 알 수 있다면 많은 정보를 얻을 수 있다. 응시 추적 기술의 발달로 응시점에 대한 정보는 다양한 응시 추적 기기에서 제공해주는 소프트웨어를 통해 얻을 수 있다. 하지만 실제 응시 깊이와 같은 정확한 정보를 추정하기란 어렵다. 응시 추적 기기를 통해 만약 실제 응시 깊이로 보정할 수 있다면 시뮬레이션, 디지털 트윈, VR 등 다양한 분야에서 현실적이고 정확한 신뢰성 있는 결과를 도출하는 것이 가능해질 것이다. 따라서 본 논문에서는 응시 추적 기기와 소프트웨어를 통해 원시 응시 깊이를 획득하고 보정하는 실험을 진행한다. 실험은 Deep Neural Network(DNN) 모델을 설계한 후 소프트웨어에서 제공하는 응시 깊이 값을 300mm에서 10,000mm까지 지정한 거리별로 획득한다. 획득한 데이터는 설계한 DNN 모델을 통해 학습을 진행하여 실제 응시 깊이와 대응하도록 보정하였다. 보정한 모델을 통해 실험을 진행한 결과, 300mm에서 10,000mm까지 지정한 거리별 297mm, 904mm, 1,485mm, 2,005mm, 3,011mm, 4,021mm, 4,972mm, 6,027mm, 7,026mm, 8,043mm, 9,021mm, 10,076mm로 실제와 비슷한 응시 깊이 값을 획득할 수 있었다.

온 디바이스 객체 검출을 위한 상황 인지 기반의 모델 선택 기법

강성주(Seongju Kang),정채은(Chaeeun Jeong),정광수(Kwangsue Chung) 한국정보과학회 2021 정보과학회 컴퓨팅의 실제 논문지 Vol.27 No.8

DNN (Deep Neural Network)은 얼굴 인식, 음성 인식, 객체 검출 등 지능형 어플리케이션에 필요한 컴퓨터 비전 기술이다. DNN모델은 수많은 은닉층과 학습 파라미터를 포함하기 때문에 객체 검출알고리즘은 많은 연산 리소스를 요구한다. 그러므로, 모바일 장치와 같이 리소스가 제한된 환경에서는 객체 검출을 수행하는데 어려움이 있다. 본 논문에서는 온 디바이스 객체 검출을 위한 상황 인지 기반의 모델 선택 기법을 제안한다. 모델의 연산 리소스를 줄이기 위하여 시공간 도메인 별 OOI (Object Of Interest) 그룹을 정의하고 각 도메인에 대해 경량화 모델을 학습시킨다. 제안하는 기법은 상황 정보를 기반으로 시공간 도메인을 결정하고 짧은 지연 시간으로 정확한 물체 감지가 가능하도록 최적의 DNN 모델을 선택한다. 기존의 객체 검출 기법과의 비교 실험을 통해 제안하는 기법이 유사한 정확도를 달성하면서 짧은 지연 시간으로 객체 검출이 가능한 것을 확인하였다. The deep neural network (DNN) is computer vision technology required for intelligent applications such as face recognition, voice recognition, and object detection, etc. Since the DNN model contains a number of hidden layers and learning parameters, the object detection algorithm requires many computational resources. Therefore, it is difficult to perform object detection in resource-constrained environments such as mobile devices. In this paper, we proposed a context-aware model selection scheme for on-device object detection. To reduce the computational resources of the DNN model, we defined the object of interest (OOI) groups for each spatiotemporal domain and trained lightweight models for each domain. The proposed scheme determines the spatiotemporal domain based on the context information and selects the optimal DNN model to enable accurate object detection with a short latency.

Deep Machine Learning and Neural Networks: An Overview

Chandrahas Mishra,Dharmendra Lal Gupta 보안공학연구지원센터 2016 International Journal of Hybrid Information Techno Vol.9 No.11

Deep learning is a technique of machine learning in artificial intelligence area. Deep learning is a refined "machine learning" algorithm that surpasses a considerable lot of its forerunners in its capacity to perceive syllables and pictures. As of now Deep learning is a greatly dynamic examination territory in machine learning and example acknowledgment society. It has increased colossal triumphs in an expansive zone of utilizations, for example, speech recognition, computer vision and natural language processing and numerous industry items. Neural networks are used to implement the machine learning or to design intelligent machines. In this paper thorough survey to all machine learning paradigms and application areas of deep machine learning and different types of neural networks with applications are discussed.

River streamflow prediction using a deep neural network: a case study on the Red River, Vietnam

Xuan-Hien Le,Hung Viet Ho,이기하 충남대학교 농업과학연구소 2019 Korean Journal of Agricultural Science Vol.46 No.4

Real-time flood prediction has an important role in significantly reducing potential damage caused by floods for urban residential areas located downstream of river basins. This paper presents an effective approach for flood forecasting based on the construction of a deep neural network (DNN) model. In addition, this research depends closely on the open-source software library, TensorFlow, which was developed by Google for machine and deep learning applications and research. The proposed model was applied to forecast the flowrate one, two, and three days in advance at the Son Tay hydrological station on the Red River, Vietnam. The input data of the model was a series of discharge data observed at five gauge stations on the Red River system, without requiring rainfall data, water levels and topographic characteristics. The research results indicate that the DNN model achieved a high performance for flood forecasting even though only a modest amount of data is required. When forecasting one and two days in advance, the Nash-Sutcliffe Efficiency (NSE) reached 0.993 and 0.938, respectively. The findings of this study suggest that the DNN model can be used to construct a real-time flood warning system on the Red River and for other river basins in Vietnam.

Dynamical prediction of two meteorological factors using the deep neural network and the long short-term memory (ΙΙ)

Shin Ki-Hong,Jung Jae-Won,Chang Ki-Ho,Kim Kyungsik,Jung Woon-Seon,Lee Dong-In,You Cheol-Hwan 한국물리학회 2022 THE JOURNAL OF THE KOREAN PHYSICAL SOCIETY Vol.80 No.12

This paper presents the predictive accuracy using two-variate meteorological factors, average temperature and average humidity, in neural network algorithms. We analyze result in fve learning architectures such as the traditional artifcial neural network, deep neural network, and extreme learning machine, long short-term memory, and long-short-term memory with peephole connections, after manipulating the computer simulation. Our neural network modes are trained on the daily time-series dataset during 7 years (from 2014 to 2020). From the trained results for 2500, 5000, and 7500 epochs, we obtain the predicted accuracies of the meteorological factors produced from outputs in ten metropolitan cities (Seoul, Daejeon, Daegu, Busan, Incheon, Gwangju, Pohang, Mokpo, Tongyeong, and Jeonju). The error statistics is found from the result of outputs, and we compare these values to each other after the manipulation of fve neural networks. As using the long-shortterm memory model in testing 1 (the average temperature predicted from the input layer with six input nodes), Tonyeong has the lowest root-mean-squared error (RMSE) value of 0.866 (%) in summer from the computer simulation to predict the temperature. To predict the humidity, the RMSE is shown the lowest value of 5.732 (%), when using the long short-term memory model in summer in Mokpo in testing 2 (the average humidity predicted from the input layer with six input nodes). Particularly, the long short-term memory model is found to be more accurate in forecasting daily levels than other neural network models in temperature and humidity forecastings. Our result may provide a computer simulation basis for the necessity of exploring and developing a novel neural network evaluation method in the future.

A Novel Image Segmentation Combined Color Recognition Algorithm through Boundary Detection and Deep Neural Network

Yu Liu,Jie Yang,Benzhen Guo,Jingjing Yang,Xiao Zhang 보안공학연구지원센터 2016 International Journal of Multimedia and Ubiquitous Vol.11 No.2

With the fast development of industry and computer science technology, the image color recognition has been a hot topic. The prior research focus more on sensor and hardware based approaches which are not intelligent or convenient. In this paper, we present a novel image segmentation combined color recognition algorithm through boundary detection and deep neural network. The deep learning algorithm can largely increase the accuracy of classification whereas cut down the processing time consumed, we adopt the deep neural network and support vector machine to extract image features both in RGB and YUV color spaces. Boundary detection in sudden change, by contrast, is more global in nature, such as texture, so need to integrate the whole information of the image. Under the guidance, we modify the current segmentation methods with boundary detection technique to serve as the pre-processing step before classifying colors. Experimental results on synthetic and real images show that the new algorithm is effective and efficient, and is relatively independent of this type of noise. Further analysis is also conducted in the final section.

심층 신경망 병렬 학습 방법 연구 동향

육동석,이효원,유인철 한국음향학회 2020 韓國音響學會誌 Vol.39 No.6

Since a large amount of training data is typically needed to train Deep Neural Networks (DNNs), a parallel training approach is required to train the DNNs. The Stochastic Gradient Descent (SGD) algorithm is one of the most widely used methods to train the DNNs. However, since the SGD is an inherently sequential process, it requires some sort of approximation schemes to parallelize the SGD algorithm. In this paper, we review various efforts on parallelizing the SGD algorithm, and analyze the computational overhead, communication overhead, and the effects of the approximations. 심층 신경망(Deep Neural Network, DNN) 모델을 대량의 학습 데이터로 학습시키기 위해서는 많은 시간이소요되기 때문에 병렬 학습 방법이 필요하다. DNN의 학습에는 일반적으로 Stochastic Gradient Descent(SGD) 방법이 사용되는데, SGD는 근본적으로 순차적인 처리가 필요하므로 병렬화하기 위해서는 다양한 근사(approximation) 방법을 적용하게 된다. 본 논문에서는 기존의 DNN 병렬 학습 알고리즘들을 소개하고 연산량, 통신량, 근사 방법 등을분석한다.

심층신경망을 활용한 비모수적 소프트웨어 신뢰성 모형과 NHPP 소프트웨어 신뢰성 성장 모형 비교

김윤수(Youn Su Kim),장인홍(In Hong Chang),이다혜(Da Hye Lee) 한국자료분석학회 2020 Journal of the Korean Data Analysis Society Vol.22 No.6

현재 소프트웨어는 다양한 기능을 수행할 수 있도록 복잡한 구성으로 만들어진다. 소프트웨어가 고장 나게 될 경우 산업 및 경제적으로 매우 큰 손실을 일으킬 수 있다. 소프트웨어에서 아주 작은 고장이 발생하더라도 큰 손실을 초래할 수 있다. 이러한 문제를 해결하기 위해 소프트웨어의 품질과 신뢰성을 추정하는 도구인 소프트웨어의 신뢰성 성장 모형이 제안되었다. 대부분의 소프트웨어 신뢰성 성장 모형은 모수적 방법인 비균질성 포아송 과정을 가정한 평균값 함수 m(t)를 기반으로 모형이 만들어지는데 이는 분포를 가정하기 때문에 일반화에 문제가 발생한다. 본 연구에서는 일반화에 문제를 해결하기 위해 본포를 가정하지 않고 데이터에 의존하는 비모수적(Non-Parametric) 방법으로 접근하여 소프트웨어 신뢰성을 측정한다. 비모수적 방법 중 인간이 사고하는 생물학적 시스템을 모태로 하는 인공신경망을 활용한다. 특히, 최근에 많이 연구되고 있는 인공신경망 중의 하나인 심층신경망을 활용하여 일반화 시킬 수 있는 소프트웨어 신뢰성 모형을 제안하고, 기존에 연구된 소프트웨어 신뢰성 성장 모형(SRGMs)과 5가지의 적합도를 비교한 후 가장 적합한 모형을 찾아 제시한다. The software is made in a complex configuration to perform various functions. If software is faulty, it can cause huge losses both industrially and economically. Even very small failures in software can lead to significant losses. To solve the problem, a software reliability growth models(SRGMs) has been proposed. Most software reliability growth models are modeled assuming a Non-Homogeneous Poison Process(NHPP). But these models has problems with generalization. In this paper, we propose a software reliability model that can be generalized using deep neural network among non-parametric methods that don’t assume distribution and depend on data. We propose more suitable model after comparing the fit of the previously studied software reliability growth models and software reliability model using deep neural network.

원어민 및 외국인 화자의 음성인식을 위한 심층 신경망 기반 음향모델링

강병옥(Kang, Byung Ok),권오욱(Kwon, Oh-Wook) 한국음성학회 2017 말소리와 음성과학 Vol.9 No.2

This paper proposes a new method to train Deep Neural Network (DNN)-based acoustic models for speech recognition of native and foreign speakers. The proposed method consists of determining multi-set state clusters with various acoustic properties, training a DNN-based acoustic model, and recognizing speech based on the model. In the proposed method, hidden nodes of DNN are shared, but output nodes are separated to accommodate different acoustic properties for native and foreign speech. In an English speech recognition task for speakers of Korean and English respectively, the proposed method is shown to slightly improve recognition accuracy compared to the conventional multi-condition training method.