• ETRI Journal
• /
• v.43 no.6
• /
• pp.1081-1092
• /
• 2021

Electroencephalography (EEG) recordings taken during the perception of music tempo contain information that estimates the tempo of a music piece. If information about this tempo stimulus in EEG recordings can be extracted and classified, it can be effectively used to construct a music-based brain-computer interface. This study proposes a novel convolutional recurrent attention model (CRAM) to extract and classify features corresponding to tempo stimuli from EEG recordings of listeners who listened with concentration to the tempo of musics. The proposed CRAM is composed of six modules, namely, network inputs, two-dimensional convolutional bidirectional gated recurrent unit-based sample encoder, sample-level intuitive attention, segment encoder, segment-level intuitive attention, and softmax layer, to effectively model spatiotemporal features and improve the classification accuracy of tempo stimuli. To evaluate the proposed method's performance, we conducted experiments on two benchmark datasets. The proposed method achieves promising results, outperforming recent methods.

• Journal of the Semiconductor & Display Technology
• /
• v.22 no.1
• /
• pp.54-58
• /
• 2023

In this study, we compared the detection accuracy of the parameters of the scaffold failure detection model. A detection algorithm based on convolutional neural network was used to construct a failure detection model for scaffold. The parameter properties of the model were changed and the results were quantitatively verified. The detection accuracy of the model for each parameter was compared and the parameter with the highest accuracy was identified. We found that the activation function has a significant impact on the detection accuracy, which is 98% for softmax.

• Annual Conference on Human and Language Technology
• /
• 2020.10a
• /
• pp.333-336
• /
• 2020

개체명 인식이란 문장에서 인명, 지명, 기관명 등과 같이 고유한 의미를 갖는 단어를 찾아 개체명을 분류하는 작업이다. 딥러닝을 활용한 연구가 수행되면서 개체명 인식에 RNN(Recurrent Neural Network)과 CRF(Condition Random Fields)를 결합한 연구가 좋은 성능을 보이고 있다. 그러나 CRF는 시간 복잡도가 분류해야 하는 클래스(Class) 개수의 제곱에 비례하고, 최근 RNN과 Softmax 모델보다 낮은 성능을 보이는 연구도 있었다. 본 논문에서는 CRF의 단점을 보완한 LAN(Label Attention Network)와 사전 학습 언어 모델인 음절 단위 ELECTRA를 활용하는 개체명 인식 모델을 제안한다.

• Proceedings of the Korean Institute of Building Construction Conference
• /
• 2022.11a
• /
• pp.245-246
• /
• 2022

Typically, the methods of crack detection on concrete structures include some problems, such as a low accuracy and expensive. To solve these problems, we proposed a neural network-based crack search method. The proposed algorithm goes through three convolutions and is classified into crack and non-crack through the softmax layer. As a result of the performance evaluation, cracks can be detected with an accuracy of 99.4 and 99.34 % at the training model and the validation model, respectively.

• Journal of Information Processing Systems
• /
• v.20 no.4
• /
• pp.458-476
• /
• 2024

To address the issue of class imbalance in network traffic data, which affects the network intrusion detection performance, a combined framework using transformers is proposed. First, Tomek Links, SMOTE, and WGAN are used to preprocess the data to solve the class-imbalance problem. Second, the transformer is used to encode traffic data to extract the correlation between network traffic. Finally, a hybrid deep learning network model combining a bidirectional gated current unit and deep neural network is proposed, which is used to extract long-dependence features. A DNN is used to extract deep level features, and softmax is used to complete classification. Experiments were conducted on the NSLKDD, UNSWNB15, and CICIDS2017 datasets, and the detection accuracy rates of the proposed model were 99.72%, 84.86%, and 99.89% on three datasets, respectively. Compared with other relatively new deep-learning network models, it effectively improved the intrusion detection performance, thereby improving the communication security of network data.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.6
• /
• pp.834-841
• /
• 2022

Recently, applying artificial intelligence technologies in various fields of production has drawn an upsurge of research interest due to the increase for smart factory and artificial intelligence technologies. A great deal of effort is being made to introduce artificial intelligence algorithms into the defect detection task. Particularly, detection of defects on the surface of metal has a higher level of research interest compared to other materials (wood, plastics, fibers, etc.). In this paper, we compare and analyze the speed and performance of defect classification by combining machine learning techniques (Support Vector Machine, Softmax Regression, Decision Tree) with dimensionality reduction algorithms (Principal Component Analysis, AutoEncoders) and two convolutional neural networks (proposed method, ResNet). To validate and compare the performance and speed of the algorithms, we have adopted two datasets ((i) public dataset, (ii) actual dataset), and on the basis of the results, the most efficient algorithm is determined.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.16 no.6
• /
• pp.526-532
• /
• 2023

Falls are fatal accidents that occur more than 420,000 times a year worldwide. Therefore, to study patients with falls, we found the association between extrinsic injury codes and principal diagnosis S-codes of patients with falls, and developed a prediction model to predict extrinsic injury codes based on the data of principal diagnosis S-codes of patients with falls. In this study, we received two years of data from 2020 and 2021 from Institution A, located in Gangneung City, Gangwon Special Self-Governing Province, and extracted only the data from W00 to W19 of the extrinsic injury codes related to falls, and developed a prediction model using W01, W10, W13, and W18 of the extrinsic injury codes of falls, which had enough principal diagnosis S-codes to develop a prediction model. 80% of the data were categorized as training data and 20% as testing data. The model was developed using MLP (Multi-Layer Perceptron) with 6 variables (gender, age, principal diagnosis S-code, surgery, hospitalization, and alcohol consumption) in the input layer, 2 hidden layers with 64 nodes, and an output layer with 4 nodes for W01, W10, W13, and W18 exogenous damage codes using the softmax activation function. As a result of the training, the first training had an accuracy of 31.2%, but the 30th training had an accuracy of 87.5%, which confirmed the association between the fall extrinsic code and the main diagnosis S code of the fall patient.

• Smart Media Journal
• /
• v.13 no.2
• /
• pp.32-43
• /
• 2024

Game agents which are behavioral agent for game playing are a crucial component of game satisfaction. However it takes a lot of time and effort to create game agents for various game levels, environments, and players. In addition, when the game environment changes such as adding contents or updating characters, new game agents need to be developed and the development difficulty gradually increases. And it is important to have a game agent that can be customized for different levels of players. This is because a game agent that can play games of various levels is more useful and can increase the satisfaction of more players than a high-level game agent. In this paper, we propose a method for learning and controlling the level of play of game agents that can be rapidly developed and fine-tuned for various game environments and changes. At this time, reinforcement learning applies a policy-based distributed reinforcement learning method IMPALA for flexible processing and fast learning of various behavioral structures. Once reinforcement learning is complete, we choose actions by sampling based on Softmax-Temperature method. From this result, we show that the game agent's play level decreases as the Temperature value increases. This shows that it is possible to easily control the play level.

• Journal of Korea Multimedia Society
• /
• v.21 no.2
• /
• pp.211-223
• /
• 2018

To understand the meaning of data is a common goal of research on unstructured data. Among these unstructured data, there are difficulties in analyzing the meaning of unstructured data related to corpus and sentences. In the existing researches, the researchers used LSA to select sentences with the most similar meaning to specific words of the sentences. However, it is problematic to examine many sentences continuously. In order to solve unstructured data classification problem, several search sites are available to classify the frequency of words and to serve to users. In this paper, we propose a method of classifying documents by using the frequency of similar words, and the frequency of non-relevant words to be applied as weights, and storing them in terms of a multi-pattern storage. We use Tensorflow's Softmax to the nearby sentences for machine learning, and utilize it for unstructured data analysis and the inference of traffic information.

• Journal of Korea Multimedia Society
• /
• v.23 no.8
• /
• pp.1019-1029
• /
• 2020

In this paper, we propose a novel knowledge distillation algorithm to create an compressed deep ensemble network coupled with the combined use of local and global features of face images. In order to transfer the capability of high-level recognition performances of the ensemble deep networks to a single deep network, the probability for class prediction, which is the softmax output of the ensemble network, is used as soft target for training a single deep network. By applying the knowledge distillation algorithm, the local feature informations obtained by training the deep ensemble network using facial subregions of the face image as input are transmitted to a single deep network to create a so-called compressed ensemble DCNN. The experimental results demonstrate that our proposed compressed ensemble deep network can maintain the recognition performance of the complex ensemble deep networks and is superior to the recognition performance of a single deep network. In addition, our proposed method can significantly reduce the storage(memory) space and execution time, compared to the conventional ensemble deep networks developed for face recognition.