Search

Your search keyword '"Kim, Jun-Seong"' showing total 116 results
Start Over Author "Kim, Jun-Seong"
116 results on '"Kim, Jun-Seong"'

1. Improving Multi-lingual Alignment Through Soft Contrastive Learning

Author

Park, Minsu, Choi, Seyeon, Choi, Chanyeol, Kim, Jun-Seong, and Sohn, Jy-yong

Subjects
Computer Science - Computation and Language
Abstract

Making decent multi-lingual sentence representations is critical to achieve high performances in cross-lingual downstream tasks. In this work, we propose a novel method to align multi-lingual embeddings based on the similarity of sentences measured by a pre-trained mono-lingual embedding model. Given translation sentence pairs, we train a multi-lingual model in a way that the similarity between cross-lingual embeddings follows the similarity of sentences measured at the mono-lingual teacher model. Our method can be considered as contrastive learning with soft labels defined as the similarity between sentences. Our experimental results on five languages show that our contrastive loss with soft labels far outperforms conventional contrastive loss with hard labels in various benchmarks for bitext mining tasks and STS tasks. In addition, our method outperforms existing multi-lingual embeddings including LaBSE, for Tatoeba dataset. The code is available at https://github.com/YAI12xLinq-B/IMASCL, Comment: 8 pages, 1 figures, Accepted at NAACL SRW 2024

Published
2024

2. Synergistic Integration of Coordinate Network and Tensorial Feature for Improving Neural Radiance Fields from Sparse Inputs

Author

Kim, Mingyu, Kim, Jun-Seong, Yun, Se-Young, and Kim, Jin-Hwa

Subjects
Computer Science - Computer Vision and Pattern Recognition, Computer Science - Artificial Intelligence
Abstract

The multi-plane representation has been highlighted for its fast training and inference across static and dynamic neural radiance fields. This approach constructs relevant features via projection onto learnable grids and interpolating adjacent vertices. However, it has limitations in capturing low-frequency details and tends to overuse parameters for low-frequency features due to its bias toward fine details, despite its multi-resolution concept. This phenomenon leads to instability and inefficiency when training poses are sparse. In this work, we propose a method that synergistically integrates multi-plane representation with a coordinate-based MLP network known for strong bias toward low-frequency signals. The coordinate-based network is responsible for capturing low-frequency details, while the multi-plane representation focuses on capturing fine-grained details. We demonstrate that using residual connections between them seamlessly preserves their own inherent properties. Additionally, the proposed progressive training scheme accelerates the disentanglement of these two features. We demonstrate empirically that our proposed method not only outperforms baseline models for both static and dynamic NeRFs with sparse inputs, but also achieves comparable results with fewer parameters., Comment: ICML2024 ; Project page is accessible at https://mingyukim87.github.io/SynergyNeRF ; Code is available at https://github.com/MingyuKim87/SynergyNeRF

Published
2024

7. Thermodynamic and Economic Analysis of Cargo Boil-Off Gas Re-Liquefaction Systems for Ammonia-Fueled LCO 2 Carriers.

Author

Kim, Jun-Seong and Kim, Do-Yeop

Subjects
LIQUEFIED gases, RANKINE cycle, WORKING fluids, HEAT exchangers, CARBON dioxide
Abstract

In this study, cargo boil-off gas (BOG) re-liquefaction systems for ammonia-fueled liquefied carbon dioxide (LCO2) carriers were analyzed. These systems use cold energy from ammonia to reliquefy the CO2 BOG. In this study, a system that can completely reliquefy the CO2 BOG at all engine loads using only one heat exchanger is proposed, instead of the existing cascade system that requires multiple components. R744, which has a low global warming potential, was used as the working fluid for the refrigeration cycle in the CO2 BOG re-liquefaction system. The organic Rankine cycle (ORC) was used to reduce the net power consumption of the system. The existing and proposed systems were classified into Case 1 (existing system), Case 2 (our proposed system), and Case 3 (Case 2 combined with an ORC). Thermodynamic and economic analyses were conducted. Case 2 is a system with a simpler configuration than Case 1, but it has a similar thermodynamic performance. Case 3 has a higher exergy destruction rate than Cases 1 and 2, owing to the ORC, but it can significantly reduce the net power consumption. The economic analysis shows that Cases 2 and 3 reduce the total annual costs by 17.4% and 20.1%, respectively, compared to Case 1. The proposed systems are significantly more advantageous for long-term operation than existing systems. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

10. Unified Parallel Intent and Slot Prediction with Cross Fusion and Slot Masking

Author

Bhasin, Anmol, Natarajan, Bharatram, Mathur, Gaurav, Jeon, Joo Hyuk, Kim, Jun-Seong, Hutchison, David, Editorial Board Member, Kanade, Takeo, Editorial Board Member, Kittler, Josef, Editorial Board Member, Kleinberg, Jon M., Editorial Board Member, Mattern, Friedemann, Editorial Board Member, Mitchell, John C., Editorial Board Member, Naor, Moni, Editorial Board Member, Pandu Rangan, C., Editorial Board Member, Steffen, Bernhard, Editorial Board Member, Terzopoulos, Demetri, Editorial Board Member, Tygar, Doug, Editorial Board Member, Goos, Gerhard, Founding Editor, Hartmanis, Juris, Founding Editor, Métais, Elisabeth, editor, Meziane, Farid, editor, Vadera, Sunil, editor, Sugumaran, Vijayan, editor, and Saraee, Mohamad, editor

Published
2019
Full Text
View/download PDF

34. Preliminary Design and Blade Optimization of a Two-Stage Radial Outflow Turbine for a CO 2 Power Cycle.

Author

Kim, Jun-Seong, Kim, You-Taek, and Kim, Do-Yeop

Subjects
*CARBON dioxide, *AXIAL flow, *RADIAL flow, *ERROR rates, *TURBINES, *ENVIRONMENTAL regulations, *ANGLES
Abstract

Recently, the CO2 power cycle has attracted attention because of tightening environmental regulations. The turbine is a factor that greatly affects the efficiency of the cycle. The radial outflow turbine is a turbomachine with the various advantages of an axial flow turbine and a radial inflow turbine, but the design theory for the turbine is uncertain. In this study, a preliminary design algorithm for a radial outflow turbine with a multi-stage configuration is presented. To verify the preliminary design algorithm, a preliminary design for a two-stage radial outflow turbine for a CO2 power cycle was carried out, and a computational fluid dynamic analysis was performed. Consequently, values close to the target performance were obtained, but blade optimization was performed to obtain more satisfactory results. The final geometry of the radial outflow turbine was obtained through optimization considering the blade exit angle related to the deviation angle, blade maximum thickness-true chord ratio, and incidence angle. In the final geometry, the error rates of power ( W ˙ ), efficiency ( η t s ), and pressure ratio ( P R t s ) between target performance and computational fluid dynamic results were improved to 5.0%, 4.8%, and 1.8%, respectively. The performance and flow characteristics of the initial and final geometries were analyzed. [ABSTRACT FROM AUTHOR]

Published
2022
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results