• 디지털융복합연구
• /
• 제11권6호
• /
• pp.199-204
• /
• 2013

음성 인식 시스템은 정확하지 않게 입력된 음성으로부터 학습 모델을 구성하고 유사한 음소 모델로 인식하기 때문에 인식률 저하를 가져온다. 따라서 본 논문에서는 바타차랴 알고리즘을 이용한 음성 인식 최적 학습 모델 구성 방법을 제안하였다. 음소가 갖는 특징을 기반으로 학습 데이터의 음소에 HMM 특징 추출 방법을 이용하였으며 유사한 학습 모델은 바타챠랴 알고리즘을 이용하여 정확한 학습 모델로 인식할 수 있도록 하였다. 바타챠랴 알고리즘을 이용하여 최적의 학습 모델을 구성하여 인식 성능을 평가하였다. 본 논문에서 제안한 시스템을 적용한 결과 음성 인식률에서 98.7%의 인식률을 나타내었다.

• Steel and Composite Structures
• /
• 제25권6호
• /
• pp.663-670
• /
• 2017

The problem of layup optimization of the composite laminates involves a very complex multidimensional solution space which is usually non-exhaustively explored using different heuristic computational methods such as genetic algorithms (GA). To ensure the convergence to the global optimum of the applied heuristic during the optimization process it is necessary to evaluate a lot of layup configurations. As a consequence the analysis of an individual layup configuration should be fast enough to maintain the convergence time range to an acceptable level. On the other hand the mechanical behavior analysis of composite laminates for any geometry and boundary condition is very convoluted and is performed by computational expensive numerical tools such as finite element analysis (FEA). In this respect some studies propose very fast FEA models used in layup optimization. However, the lower bound of the execution time of FEA models is determined by the global linear system solving which in some complex applications can be unacceptable. Moreover, in some situation it may be highly preferred to decrease the optimization time with the cost of a small reduction in the analysis accuracy. In this paper we explore some machine learning techniques in order to estimate the failure of a layup configuration. The estimated response can be qualitative (the configuration fails or not) or quantitative (the value of the failure factor). The procedure consists of generating a population of random observations (configurations) spread across solution space and evaluating using a FEA model. The machine learning method is then trained using this population and the trained model is then used to estimate failure in the optimization process. The results obtained are very promising as illustrated with an example where the misclassification rate of the qualitative response is smaller than 2%.

• Nuclear Engineering and Technology
• /
• 제54권8호
• /
• pp.3027-3033
• /
• 2022

Background: In this study, various types of deep-learning models for predicting in vitro radiosensitivity from gene-expression profiling were compared. Methods: The clonogenic surviving fractions at 2 Gy from previous publications and microarray gene-expression data from the National Cancer Institute-60 cell lines were used to measure the radiosensitivity. Seven different prediction models including three distinct multi-layered perceptrons (MLP), four different convolutional neural networks (CNN) were compared. Folded cross-validation was applied to train and evaluate model performance. The criteria for correct prediction were absolute error < 0.02 or relative error < 10%. The models were compared in terms of prediction accuracy, training time per epoch, training fluctuations, and required calculation resources. Results: The strength of MLP-based models was their fast initial convergence and short training time per epoch. They represented significantly different prediction accuracy depending on the model configuration. The CNN-based models showed relatively high prediction accuracy, low training fluctuations, and a relatively small increase in the memory requirement as the model deepens. Conclusion: Our findings suggest that a CNN-based model with moderate depth would be appropriate when the prediction accuracy is important, and a shallow MLP-based model can be recommended when either the training resources or time are limited.

• Smart Structures and Systems
• /
• 제24권2호
• /
• pp.233-242
• /
• 2019

In this paper, an inverse approach based on uniform load surface (ULS) is presented for structural damage localization and quantification. The ULS is excellent approximation for deformed configuration of a structure under distributed unit force applied on all degrees of freedom. The ULS make use of natural frequencies and mode shapes of structure and in mathematical point of view is a weighted average of mode shapes. An objective function presented to damage detection is discrepancy between the ULS of monitored structure and numerical model of structure. Solving this objective function to find minimum value yields damage's parameters detection. The teaching-learning based optimization algorithm has been employed to solve inverse problem. The efficiency of present damage detection method is demonstrated through three numerical examples. By comparison between proposed objective function and another objective function which make use of natural frequencies and mode shapes, it is revealed present objective function have faster convergence and is more sensitive to damage. The method has good robustness against measurement noise and could detect damage by using the first few mode shapes. The results indicate that the proposed method is reliable technique to damage detection in structures.

• International journal of advanced smart convergence
• /
• 제13권2호
• /
• pp.80-87
• /
• 2024

This paper focuses on improving accuracy in constrained computing settings by employing the ReLU (Rectified Linear Unit) activation function. The research conducted involves modifying parameters of the ReLU function and comparing performance in terms of accuracy and computational time. This paper specifically focuses on optimizing ReLU in the context of a Multilayer Perceptron (MLP) by determining the ideal values for features such as the dimensions of the linear layers and the learning rate (Ir). In order to optimize performance, the paper experiments with adjusting parameters like the size dimensions of linear layers and Ir values to induce the best performance outcomes. The experimental results show that using ReLU alone yielded the highest accuracy of 96.7% when the dimension sizes were 30 - 10 and the Ir value was 1. When combining ReLU with the Adam optimizer, the optimal model configuration had dimension sizes of 60 - 40 - 10, and an Ir value of 0.001, which resulted in the highest accuracy of 97.07%.

• International journal of advanced smart convergence
• /
• 제6권4호
• /
• pp.73-79
• /
• 2017

The Convolutional Neural Network (CNN) has shown an excellent performance in computer vision task. Applications of CNN include image classification, object detection in images, autonomous driving, etc. This paper will evaluate the performance of CNN model with ReLU and SELU as activation function. The evaluation will be performed on four different choices of hyperparameter which are initialization method, network configuration, optimization technique, and regularization. We did experiment on each choice of hyperparameter and show how it influences the network convergence and test accuracy. In this experiment, we also discover performance improvement when using SELU as activation function over ReLU.

• International Journal of Computer Science & Network Security
• /
• 제23권11호
• /
• pp.99-109
• /
• 2023

The darknet is frequently referred to as the hub of illicit online activity. In order to keep track of real-time applications and activities taking place on Darknet, traffic on that network must be analysed. It is without a doubt important to recognise network traffic tied to an unused Internet address in order to spot and investigate malicious online activity. Any observed network traffic is the result of mis-configuration from faked source addresses and another methods that monitor the unused space address because there are no genuine devices or hosts in an unused address block. Digital systems can now detect and identify darknet activity on their own thanks to recent advances in artificial intelligence. In this paper, offer a generalised method for deep learning-based detection and classification of darknet traffic. Furthermore, analyse a cutting-edge complicated dataset that contains a lot of information about darknet traffic. Next, examine various feature selection strategies to choose a best attribute for detecting and classifying darknet traffic. For the purpose of identifying threats using network properties acquired from darknet traffic, devised a hybrid deep learning (DL) approach that combines Recurrent Neural Network (RNN) and Bidirectional LSTM (BiLSTM). This probing technique can tell malicious traffic from legitimate traffic. The results show that the suggested strategy works better than the existing ways by producing the highest level of accuracy for categorising darknet traffic using the Black widow optimization algorithm as a feature selection approach and RNN-BiLSTM as a recognition model.

• Computers and Concrete
• /
• 제32권2호
• /
• pp.149-163
• /
• 2023

One of the main design parameters traditionally utilized in projects of geotechnical engineering is the uniaxial compressive strength. The present paper employed three artificial intelligence methods, i.e., the stochastic fractal search (SFS), the multi-verse optimization (MVO), and the vortex search algorithm (VSA), in order to determine the compressive strength of concrete (CSC). For the same reason, 1030 concrete specimens were subjected to compressive strength tests. According to the obtained laboratory results, the fly ash, cement, water, slag, coarse aggregates, fine aggregates, and SP were subjected to tests as the input parameters of the model in order to decide the optimum input configuration for the estimation of the compressive strength. The performance was evaluated by employing three criteria, i.e., the root mean square error (RMSE), mean absolute error (MAE), and the determination coefficient (R²). The evaluation of the error criteria and the determination coefficient obtained from the above three techniques indicates that the SFS-MLP technique outperformed the MVO-MLP and VSA-MLP methods. The developed artificial neural network models exhibit higher amounts of errors and lower correlation coefficients in comparison with other models. Nonetheless, the use of the stochastic fractal search algorithm has resulted in considerable enhancement in precision and accuracy of the evaluations conducted through the artificial neural network and has enhanced its performance. According to the results, the utilized SFS-MLP technique showed a better performance in the estimation of the compressive strength of concrete (R²=0.99932 and 0.99942, and RMSE=0.32611 and 0.24922). The novelty of our study is the use of a large dataset composed of 1030 entries and optimization of the learning scheme of the neural prediction model via a data distribution of a 20:80 testing-to-training ratio.

• Journal of Multimedia Information System
• /
• 제8권3호
• /
• pp.147-158
• /
• 2021

Versatile Video Coding (VVC) is the latest video coding standard developed by Joint Video Exploration Team (JVET). In VVC, the quadtree plus multi-type tree (QT+MTT) structure of coding unit (CU) partition is adopted, and its computational complexity is considerably high due to the brute-force search for recursive rate-distortion (RD) optimization. In this paper, we aim to reduce the time complexity of inter-picture prediction mode since the inter prediction accounts for a large portion of the total encoding time. The problem can be defined as classifying the split mode of each CU. To classify the split mode effectively, a novel convolutional neural network (CNN) called multi-level tree (MLT-CNN) architecture is introduced. For boosting classification performance, we utilize additional information including inter-picture information while training the CNN. The overall algorithm including the MLT-CNN inference process is implemented on VVC Test Model (VTM) 11.0. The CUs of size 128×128 can be the inputs of the CNN. The sequences are encoded at the random access (RA) configuration with five QP values {22, 27, 32, 37, 42}. The experimental results show that the proposed algorithm can reduce the computational complexity by 11.53% on average, and 26.14% for the maximum with an average 1.01% of the increase in Bjøntegaard delta bit rate (BDBR). Especially, the proposed method shows higher performance on the sequences of the A and B classes, reducing 9.81%~26.14% of encoding time with 0.95%~3.28% of the BDBR increase.

• 정보처리학회논문지:컴퓨터 및 통신 시스템
• /
• 제7권1호
• /
• pp.9-18
• /
• 2018

분산 처리 플랫폼에서 다양한 빅 데이터 처리 어플리케이션들의 수행 성능 향상에 대한 관심이 높아지고 있다. 이에 따라 범용적인 분산 처리 플랫폼인 아파치 스파크에서 어플리케이션들의 처리 성능 최적화에 대한 연구들이 활발하게 진행되고 있다. 스파크에서 데이터 처리 어플리케이션들의 수행 성능을 향상시키기 위해서는 스파크의 분산처리모델인 Directed Acyclic Graph(DAG)에 알맞은 형태로 어플리케이션을 최적화시켜야 하고 어플리케이션의 처리 특징을 고려하여 스파크 시스템 파라미터들을 설정해야 하기 때문에 매우 어렵다. 기존 연구들은 각각의 어플리케이션의 처리 성능에 영향을 주는 하나의 요소에 대한 부분적인 연구를 수행했고, 최종적으로 어플리케이션의 성능개선을 이뤄냈지만 스파크의 전반적인 처리과정을 고려한 성능 최적화를 다루지 않았을 뿐만 아니라 처리성능과 상관관계를 갖는 다양한 요소들의 복합적인 상호작용을 고려하지 못했다. 본 연구에서는 스파크에서 일반적인 데이터 처리 어플리케이션의 수행 과정을 분석하고, 분석된 결과를 토대로 어플리케이션의 처리과정 중 스테이지 내부와 스테이지 사이에서 성능 향상을 위한 처리 전략을 제안한다. 또한 스파크의 시스템 설정 파라미터 중 분산 병렬처리와 밀접한 관계를 갖는 파티션 병렬화에 따른 어플리케이션의 수행성능을 분석하고 적합한 파티셔닝 최적화 기법을 제안한다. 3가지 성능 향상 전략의 실효성을 입증하기 위해 일반적인 데이터 처리 어플리케이션: WordCount, Pagerank, Kmeans에 각각의 방법을 사용했을 때의 성능 향상률을 제시한다. 또한 제안한 3가지 성능 최적화 기법들이 함께 적용될 때 복합적인 성능향상 시너지를 내는지를 확인하기 위해 모든 기법들이 적용됐을 때의 성능 향상률을 제시함으로써 본 연구에서 제시하는 전략들의 실효성을 입증한다.