Digital Library[ Search Result ]
Prediction of Dehydrogenation Enthalpy Using Graph Isomorphism Network
Kun Young Choi, Woo Hyun Yuk, Jeong Woo Han, Cham Kill Hong
http://doi.org/10.5626/JOK.2024.51.5.406
This paper conducts dehydrogenation enthalpy prediction that could play an important role in selecting optimal liquid organic hydrogen carriers. We employed graph convolutional networks, which produced molecular embeddings for the prediction. Specifically, we adopted Graph Isomorphism Network (GIN) known to be the most expressive graph-based representation learning model. Our proposed approach could build molecular embeddings. Our proposed approach outperformed conventional machine learning solutions and traditional representations based on chemical physics algorithms. In addition, the performance of the proposed model could be improved with small batch sizes and deeper GCN layers using skip connections.
ConTL: Improving the Performance of EEG-based Emotion Recognition via the Incorporation of CNN, Transformer and LSTM
Hyunwook Kang, Byung Hyung Kim
http://doi.org/10.5626/JOK.2024.51.5.454
This paper proposes a hybrid-network called ConTL, which is composed of a convolutional neural network (CNN), Transformer, and long short-term memory (LSTM) for EEG-based emotion recognition. Firstly, CNN is exploited to learn local features from the input EEG signals. Then, the Transformer learns global temporal dependencies from the output features. To further learn sequential dependencies of the time domain, the output features from the Transformer are fed to the bi-directional LSTM. To verify the effects of the proposed model, we compared the classification accuracies with five state-of-the-art models. There was an 0.73% improvement on SEED-IV compared to CCNN, and improvements of 0.97% and 0.63% were observed compared to DGCNN for valence and arousal of DEAP, respectively.
Graph Structure Learning-Based Neural Network for ETF Price Movement Prediction
Hyeonsoo Jo, Jin-gee Kim, Taehun Kim, Kijung Shin
http://doi.org/10.5626/JOK.2024.51.5.473
Exchange-Traded Funds (ETFs) are index funds that mirror particular market indices, usually associated with their low risk and expense ratio to individual investors. Various methods have emerged for accurately predicting ETF price movements, and recently, AI-based technologies have been developed. One representative method involves using time-series-based neural networks to predict the price movement of ETFs. This approach effectively incorporates past price information of ETFs, allowing the prediction of their movement. However, it has a limitation as it only utilizes historical information of individual ETFs and does not account for the relationships and interactions between different ETFs. To address this issue, we propose a model that can capture relationships between ETFs. The proposed model uses graph structure learning to infer a graph representing relationships between ETFs. Based on this, a graph neural network predicts the ETF price movement. The proposed model demonstrates superior performance compared to time-series-based deep-learning models that only use individual ETF information.
An Image Harmonization Method with Improved Visual Uniformity of Composite Images in Various Lighting Colors
Doyeon Kim, Jonghwa Shim, Hyeonwoo Kim, Changsu Kim, Eenjun Hwang
http://doi.org/10.5626/JOK.2024.51.4.345
Image composition is a technique that creates a composite image by arranging foreground objects extracted from other images onto a background image. To improve the visual uniformity of the composite images, deep learning-based image harmonization techniques that adjust the lighting and color of foreground objects to match the background image have been actively proposed recently. However, existing techniques have limited performance in visual uniformity because they adjust colors only for the lighting color distribution of the dataset used for training. To address this problem, we propose a novel image harmonization scheme that has robust performance for various lighting colors. First, iHColor, a new dataset composed of various lighting color distributions, is built through data preprocessing. Then, a pre-trained GAN-based Harmonization model is fine-tuned using the iHColor dataset. Through experiments, we demonstrate that the proposed scheme can generate harmonized images with better visual uniformity than existing models for various lighting colors.
Graph Structure Learning: Reflecting Types of Relationships between Sensors in Multivariate Time Series Anomaly Detection
http://doi.org/10.5626/JOK.2024.51.3.236
Sensors are used to monitor systems in various fields, such as water treatment systems and smart factories. Anomalies in the system can be detected by analyzing multivariate time series consisting of sensor data. To efficiently detect anomalies, information about the relationships between sensors is required, but this information is generally difficult to obtain. To solve this problem, the previous work used sensor data to identify relationships between sensors, which were then represented using a graph structure. However, in this process, the graph structure only reflects the presence of relationships between sensors, not the types of relationships between sensors. In this pap er, we considered the types of relationships between sensors in graph structure learning and analyzed multivariate time series to detect anomalies in the system. Experiments show that improving detection accuracy in graph structure learning for multivariate time series anomaly detection involves taking into account the different kinds of relationships among sensors.
A Hybrid Deep Learning Model for Real-Time Forecasting Fire Temperature in Underground Utility Tunnel Based on Residual CNN-LSTM
http://doi.org/10.5626/JOK.2024.51.2.131
Underground utility tunnels (UUTs) play major roles in sustaining the life of citizens and industries with regard to carrying electricity, telecommunication, water supply pipes. Fire is one of the most commonly common disasters in underground facilities, which can be prevented through proper management. This paper proposes a hybrid deep learning model named Residual CNN-LSTM to predict fire temperatures. Scenarios of underground facility fire outbreaks were created and fire temperature data was collected using FDS software. In the experiment, we analyzed the appropriate depth of residual learning of the proposed model and compared the performance to other predictive models. The results showed that RMSE, MAE and MAPE of Residual CNN-LSTM are each 0.061529, 0.053851, 6.007076 respectively, making Residual CNN-LSTM far superior to other models in terms of its predictive performance.
Explainable Artificial Intelligence in Molecular Graph Classification
Yeongyeong Son, Yewon Shin, Sunyoung Kwon
http://doi.org/10.5626/JOK.2024.51.2.157
With the advancement of artificial intelligence (AI), there is a growing need for explainable artificial intelligence (XAI). Recently, Graph neural network-based XAI research has been actively conducted, but it mainly focuses on generic graphs. Due to the distinctive characteristics relying on the chemical properties of molecular graphs, we emphasize the necessity for research to investigate whether existing XAI techniques can provide interpretability in molecular graphs. In this paper, we employ existing XAI techniques to molecular graphs and assess them quantitatively and qualitatively to see their interpretability. Furthermore, we examine the outcomes after standardizing the significance ratio of essential features, highlighting the significance of sparsity as one of the XAI evaluation metrics.
Polyphonic Music Generation with Sequence Generative Adversarial Networks
Sang-gil Lee, Uiwon Hwang, Seonwoo Min, Sungroh Yoon
http://doi.org/10.5626/JOK.2024.51.1.78
In this paper, we propose an application of sequence generative adversarial networks (SeqGAN) for generating polyphonic musical sequences. We introduce a representation of polyphonic MIDI files that could encapsulate both chords and melodies with dynamic timings. This method condensed the duration, octaves, and keys of both melodies and chords into a single word vector representation. Our generator composed of recurrent neural networks was trained to predict distributions of musical word sequences. Additionally, we employed the least square loss function for the discriminator to stabilize training of the model. Our model could create sequences that are musically coherent. It exhibited improved quantitative and qualitative measures.
C3DSG: A 3D Scene Graph Generation Model Using Point Clouds of Indoor Environment
http://doi.org/10.5626/JOK.2023.50.9.758
To design an effective deep neural network model to generate 3D scene graphs from point clouds, the following three challenging issues need to be resolved: 1) to decide how to extract effective geometric features from point clouds, 2) to determine what non-geometric features are used complementarily for recognizing 3D spatial relationships between two objects, and 3) to decide which spatial reasoning mechanism is used. To address these challenging issues, we proposed a novel deep neural network model for generating 3D scene graphs from point clouds of indoor environments. The proposed model uses both geometric features of 3D point cloud extracted using Point Transformer and various non-geometric features such as linguistic features and relative comparison features that can help predict the 3D spatial relationship between objects. In addition, the proposed model uses a new NE-GAT graph neural network module that can apply attention to both object nodes and edges connecting them to effectively derive spatial context between objects. Conducting a variety of experiments using 3DSSG benchmark dataset, effectiveness and superiority of the proposed mode were proven.
Applying Deep Neural Networks and Random Forests to Predict the Pathogenicity of Single Nucleotide Variants in Hereditary Cancer-associated Genes
Da-Bin Lee, Seonhwa Kim, Moonjong Kang, Changbum Hong, Kyu-Baek Hwang
http://doi.org/10.5626/JOK.2023.50.9.746
The recent proliferation of genetic testing has made it possible to explore an individual"s genetic variants and use pathogenicity information to diagnose and prevent genetic diseases. However, the number of identified variants with pathogenicity information is quite small. A method for predicting the pathogenicity of variants by machine learning was proposed to address this problem. In this study, we apply and compare deep neural networks with random forests and logistic regression, which have been widely used in previous studies, to predict variant pathogenicity. The experimental data consisted of 1,068 single-nucleotide variants in genes associated with hereditary cancers. Experiments on 100 random data-sets generated for hyperparameter selection showed that random forests performed best in terms of area under the precision-recall curve. On 15 holdout gene data-sets, deep neural networks performed best on average, but the difference in performance from the second-best random forest was not significant. Logistic regression was also statistically significantly worse than that of either model. In conclusion, we found that deep neural networks and random forests were generally better than logistic regression at predicting the pathogenicity of single-nucleotide variants associated with hereditary cancer.
Search
Journal of KIISE
- ISSN : 2383-630X(Print)
- ISSN : 2383-6296(Electronic)
- KCI Accredited Journal
Editorial Office
- Tel. +82-2-588-9240
- Fax. +82-2-521-1352
- E-mail. chwoo@kiise.or.kr