Vol. 51, No. 6,
Jun. 2024
Digital Library
Running IO500 Benchmark and Applying Optimal Parameters on Lustre-based Storage Systems
Hyosil Kim, Byoungjun Seo, Sejin Hwang, Juyeun Han
http://doi.org/10.5626/JOK.2024.51.6.483
In this paper, a comparative study is conducted by executing IO500 benchmarks with various tuning factors in the Lustre-based storage environment, where Lustre is a distributed parallel file system. Contributions of this paper are as follows: 1) enhancing user understanding of IO500 benchmarks and Lustre file system, and 2) establishing a large-scale Lustre environment to analyze factors that can enhance IO500 performance from three perspectives, and 3) presenting results of applied enhancements through experimentation. Experiments were conducted concerning Data-on-MDT (DoM), MPI process count (NP), and MPI CPU affinity parameters. We demonstrated that the performance was improved when DoM was configured, an appropriate number of MPI processes was set, and the '--bind-to core-overload-allowed' option was employed for CPU affinity.
A Hybrid Deep Learning Model for Generating Time-series Fire Data in Underground Utility Tunnel based on Convolutional Attention TimeGAN
http://doi.org/10.5626/JOK.2024.51.6.490
Underground utility tunnels (UUTs) play a crucial role in urban operation and management. Fires are the most common disasters in the facilities, and there is a growing demand for fire management systems using artificial intelligence (AI). However, due to the difficulty of collecting fire data for AI training, utilizing data generation models reflecting the key characteristics of real fires can be an alternative. In this paper, we propose an approach for generating AI training data based on the fire data generation model CA-TimeGAN. To collect fire simulation data for training the proposed model, we constructed a UUT in Chungbuk Ochang within the fire dynamic simulator (FDS) virtual environment. In the experiments, we compared data generated by TimeGAN and CA-TimeGAN, verifying the data quality and effectiveness. Discriminative score converged to 0.5 for both CA-TimeGAN and TimeGAN. Predictive scores improved by 66.1% compared to models trained only on simulated data and by 22.9% compared to models incorporating TimeGAN-generated data. PCA and t-SNE analyses showed that the distribution of generated data was similar to that of simulated data.
Drug-Drug Interaction Prediction Model Based on Deep Learning Using Drug Information Document Embedding
http://doi.org/10.5626/JOK.2024.51.6.503
The use of polypharmacy has emerged as a promising approach for various diseases, including cancer, hypertension, and asthma. However, the use of polypharmacy can result in unexpected interactions, which may lead to adverse drug effects. Therefore, predicting drug-drug interactions (DDI) is essential for safe medication practices. In this study, we propose a drug-drug interaction prediction model based on deep learning using document embedding to represent the drug. We generate documents about drug information by combining DrugBank data, which includes drug descriptions, indications, mechanisms of action, pharmacodynamics, and toxicity. Then we use Doc2Vec and BioSentVec language models to generate drug representation vectors from the drug information documents. The two drug vectors are paired and input into the deep learning-based prediction model, which outputs the likelihood of interaction between the two drugs. Our goal is to construct the optimal model for predicting drug-drug interactions by comparing the performance under various conditions, including language embedding model performance and adjustments for data imbalance. We expect the proposed model to be utilized for the advanced prediction of drug interactions during the drug prescription process and the clinical stages of new drug development.
Creating a of Noisy Environment Speech Mixture Dataset for Korean Speech Separation
Jaehoo Jang, Kun Park, Jeongpil Lee, Myoung-Wan Koo
http://doi.org/10.5626/JOK.2024.51.6.513
In the field of speech separation, models are typically trained using datasets that contain mixtures of speech and overlapping noise. Although there are established international datasets for advancing speech separation techniques, Korea currently lacks a similar precedent for constructing datasets with overlapping speech and noise. Therefore, this paper presents a dataset generator specifically designed for single-channel speech separation models tailored to the Korean language. The Korean Speech mixture with Noise dataset is introduced, which has been constructed using this generator. In our experiments, we train and evaluate a Conv-TasNet speech separation model using the newly created dataset. Additionally, we verify the dataset's efficacy by comparing the Character Error Rate (CER) between the separated speech and the original speech using a pre-trained speech recognition model.
An Automated Interior Design Model using Interior Design Guidelines and Proximal Policy Optimization
Chanyoung Yoon, Soobin Yim, Sangbong Yoo, Yun Jang
http://doi.org/10.5626/JOK.2024.51.6.519
The interior design of a residential space greatly influences the satisfaction and impression of its residents. However, interior design is not easily accessible due to its requirement for professional design knowledge. Therefore, optimization and deep learning methods for automated interior design have been proposed. Nevertheless, these technologies have encountered difficulties such as taking a considerable amount of time to solve problems or requiring extensive training data. In this paper, we propose an automated interior design model using deep reinforcement learning. In reinforcement learning, there is no need to obtain training data because the agent learns a policy that interacts with the environment and maximizes the cumulative reward. We designed interior design guidelines proposed in previous studies as a reward function to create interior layouts that satisfy functional and visual criteria. Reinforcement learning agents used PPO to arrange furniture in continuous positions. We evaluated the performance of the proposed model through two experiments: a reward comparison experiment based on different combinations of furniture and room shapes, and a design comparison experiment based on different combinations of reward functions.
Developement of SITL & HITL System based on PX4-Matlab for VTOL Test
Donghyeon Ko, Minkyu Kim, Jinseok Jung, SungTae Moon
http://doi.org/10.5626/JOK.2024.51.6.528
PX4, the open-source-based flight control computer, is garnering significant attention for its ability to support various drone configurations. The integration of Gazebo and PX4 enables the validation of flight control algorithms in simulations, making it widely utilized in system development. However, when developing specialized drone configurations not natively supported by Gazebo, it becomes challenging to create vehicle models. In addition the fidelity of the drone component models provided by Gazebo is often not high, leading to discrepancies between simulation and actual aircraft operation results. In the aerospace field, Matlab/Simulink, known for its high-fidelity drone models, has been widely used. However, the integration of Matlab/Simulink with PX4 has presented challenges, resulting in the need to maintain separate source code for simulation and real-world operations, leading to duplicated development efforts. This paper proposes a PX4-Matlab simulator, which leverages Matlab/Simulink commonly used in the drone industry, as an alternative to the conventional approach of using PX4 and Gazebo. To verify the proposed system, it was applied and tested on the LC-62 VTOL drone.
Improvement of Background Inpainting using Binary Masking of a Generated Image
Jihoon Lee, Chan Ho Bae, Seunghun Lee, Myung-Seok Choi, Ryong Lee, Sangtae Ahn
http://doi.org/10.5626/JOK.2024.51.6.537
Recently, image generation technology has been rapidly advancing in the field of deep learning. One of the most effective ways to represent images is by using text prompts to generate them. The performance of models that generate images using this technique is outstanding. However, it is not easy to naturally change specific parts of an image using only text prompts. This is considered a typical problem with conventional image generation models. Thus, in this study, we developed a background inpainting technique that extracts text for each area of an image and uses it as a basis to seamlessly change the background while preserving the objects in the image. In particular, the background transformation inpainting technique developed in this study has the advantage of not only transforming a single image but also rapidly transforming multiple images. Therefore, the proposed text prompt-based image style transfer can be used in fields with limited data for training, and the technique could enhance the performance of models through image augmentation.
Unified Prediction of Pedestrian Intention to Jaywalk Based on Parallel Deep Learning Scheme
http://doi.org/10.5626/JOK.2024.51.6.545
Urbanization has led to diversification in traffic accidents and parking issues, with pedestrian accidents at crosswalks accounting for over 30% of traffic fatalities. Particularly concerning are situations where pedestrians are not anticipated by drivers during red signal conditions, as the potential for severe injuries is high. To address this issue, we propose a deep learning-based integrated pedestrian crossing intent prediction system. The system uses the YOLOv5 object detection model to identify pedestrian actions that indicate crossing intent. At the same time, it utilzes the MMPOSE joint prediction model to classify the pedestrian's perspective. By analyzing pedestrian actions, perspectives, and the distance between the pedestrian and the crosswalk, the system predicts crossing intent in various scenarios. Future research based on this study is expected to contribute to diverse application studies aimed at enhancing traffic safety in autonomous driving.
Development of Personalized Autonomous Driving Agents Using Imitation Learning
Ji Hye Ok, Wookyoung Kim, Honguk Woo
http://doi.org/10.5626/JOK.2024.51.6.558
The rise of Autonomous Vehicles (AVs) has brought humans and robots together on the same roads. As AVs integrate into the existing road system, it is crucial for them to establish a connection with human drivers and operate in a way that is convenient to humans. Moreover, as the desire for personalized autonomous driving experiences frows, there is a need to meet the demand for ‘personalized’ AVs. This paper examines imitation learning methods that imitate the driving behaviors of rule-based agents. It also proposes a controlled multi-objective imitation learning approach to generate diverse driving policies based on given data. Additionally, the study assesses the derived policies in various scenarios using the Carla simulator.
A Token Selection Method for Effective Token Pruning in Vision Transformers
http://doi.org/10.5626/JOK.2024.51.6.567
The self-attention-based models, vision transformers, have recently been employed in the field of computer vision. While achieving excellent performance in a variety of tasks, the computation costs increase in proportion to the number of tokens during inference, which causes a degradation in inference speed. Especially when deploying the model in real-world scenarios, many limitations could be encountered. To address this issue, we propose a new token importance measurement, which can be obtained by modifying the structure of multi-head self-attention in vision transformers. By pruning less important tokens through our method during inference, we can improve inference speed while preserving performance. Furthermore, our proposed method, which requires no additional parameters, exhibits better robustness without fine-tuning and demonstrates that it can maximize performance when integrated with existing token pruning methods.
Prediction of Cancer Prognosis Using Patient-Specific Cancer Driver Gene Information
http://doi.org/10.5626/JOK.2024.51.6.574
Accurate prediction of cancer prognosis is crucial for effective treatment. Consequently, numerous studies on cancer prognosis have been conducted, with recent research leveraging various machine learning techniques such as deep learning. In this paper, we first constructed patient-specific gene networks for each patient, then selected patient-specific cancer driver genes, considering the heterogeneity of cancer. We propose a deep neural architecture that can predict the prognosis more accurately using patient-specific cancer driver gene information. When our method was applied to gene expression data for 11 types of cancer, it demonstrated a significantly higher prediction accuracy compared to the existing methods.
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