Search

Your search keyword '"Lee, Suho"' showing total 232 results
Start Over Author "Lee, Suho"
232 results on '"Lee, Suho"'

2. StochCA: A Novel Approach for Exploiting Pretrained Models with Cross-Attention

Author

Seo, Seungwon, Lee, Suho, and Hwang, Sangheum

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Utilizing large-scale pretrained models is a well-known strategy to enhance performance on various target tasks. It is typically achieved through fine-tuning pretrained models on target tasks. However, na\"{\i}ve fine-tuning may not fully leverage knowledge embedded in pretrained models. In this study, we introduce a novel fine-tuning method, called stochastic cross-attention (StochCA), specific to Transformer architectures. This method modifies the Transformer's self-attention mechanism to selectively utilize knowledge from pretrained models during fine-tuning. Specifically, in each block, instead of self-attention, cross-attention is performed stochastically according to the predefined probability, where keys and values are extracted from the corresponding block of a pretrained model. By doing so, queries and channel-mixing multi-layer perceptron layers of a target model are fine-tuned to target tasks to learn how to effectively exploit rich representations of pretrained models. To verify the effectiveness of StochCA, extensive experiments are conducted on benchmarks in the areas of transfer learning and domain generalization, where the exploitation of pretrained models is critical. Our experimental results show the superiority of StochCA over state-of-the-art approaches in both areas. Furthermore, we demonstrate that StochCA is complementary to existing approaches, i.e., it can be combined with them to further improve performance. Our code is available at https://github.com/daintlab/stochastic_cross_attention, Comment: The updated version was published in Neural Networks (https://www.sciencedirect.com/science/article/abs/pii/S0893608024005872). The first two authors contributed equally

Published
2024
Full Text
View/download PDF

5. Integrative analysis of spatial and single-cell transcriptome data from human pancreatic cancer reveals an intermediate cancer cell population associated with poor prognosis

Author

Kim, Seongryong, Leem, Galam, Choi, Junjeong, Koh, Yongjun, Lee, Suho, Nam, Sang-Hee, Kim, Jin Su, Park, Chan Hee, Hwang, Ho Kyoung, Min, Kyoung Il, Jo, Jung Hyun, Lee, Hee Seung, Chung, Moon Jae, Park, Jeong Youp, Park, Seung Woo, Song, Si Young, Shin, Eui-Cheol, Kang, Chang Moo, Bang, Seungmin, and Park, Jong-Eun

Published
2024
Full Text
View/download PDF

6. Few-shot Fine-tuning is All You Need for Source-free Domain Adaptation

Author

Lee, Suho, Seo, Seungwon, Kim, Jihyo, Lee, Yejin, and Hwang, Sangheum

Subjects
Computer Science - Computer Vision and Pattern Recognition, Computer Science - Machine Learning
Abstract

Recently, source-free unsupervised domain adaptation (SFUDA) has emerged as a more practical and feasible approach compared to unsupervised domain adaptation (UDA) which assumes that labeled source data are always accessible. However, significant limitations associated with SFUDA approaches are often overlooked, which limits their practicality in real-world applications. These limitations include a lack of principled ways to determine optimal hyperparameters and performance degradation when the unlabeled target data fail to meet certain requirements such as a closed-set and identical label distribution to the source data. All these limitations stem from the fact that SFUDA entirely relies on unlabeled target data. We empirically demonstrate the limitations of existing SFUDA methods in real-world scenarios including out-of-distribution and label distribution shifts in target data, and verify that none of these methods can be safely applied to real-world settings. Based on our experimental results, we claim that fine-tuning a source pretrained model with a few labeled data (e.g., 1- or 3-shot) is a practical and reliable solution to circumvent the limitations of SFUDA. Contrary to common belief, we find that carefully fine-tuned models do not suffer from overfitting even when trained with only a few labeled data, and also show little change in performance due to sampling bias. Our experimental results on various domain adaptation benchmarks demonstrate that the few-shot fine-tuning approach performs comparatively under the standard SFUDA settings, and outperforms comparison methods under realistic scenarios. Our code is available at https://github.com/daintlab/fewshot-SFDA ., Comment: The first two authors contributed equally

Published
2023

7. Author Correction: SLC6A20 transporter: a novel regulator of brain glycine homeostasis and NMDAR function

Author

Bae, Mihyun, Roh, Junyeop Daniel, Kim, Youjoung, Kim, Seong Soon, Han, Hye Min, Yang, Esther, Kang, Hyojin, Lee, Suho, Kim, Jin Yong, Kang, Ryeonghwa, Jung, Hwajin, Yoo, Taesun, Kim, Hyosang, Kim, Doyoun, Oh, Heejeong, Han, Sungwook, Kim, Dayeon, Han, Jinju, Bae, Yong Chul, Kim, Hyun, Ahn, Sunjoo, Chan, Andrew M, Lee, Daeyoup, Kim, Jin Woo, and Kim, Eunjoon

Published
2024
Full Text
View/download PDF

8. Excitatory synapses and gap junctions cooperate to improve Pv neuronal burst firing and cortical social cognition in Shank2-mutant mice.

Author

Lee, Taekyung, Shin, Hyogeun, Cho, Il-Joo, Deisseroth, Karl, Kim, Sang, Park, Joo, Jung, Min, Paik, Se-Bum, Kim, Eunjoon, Lee, Eunee, Lee, Seungjoon, Shin, Jae, Choi, Woochul, Chung, Changuk, Lee, Suho, Kim, Jihye, Ha, Seungmin, Kim, Ryunhee, Yoo, Taesun, Yoo, Ye-Eun, Kim, Jisoo, Noh, Young, Rhim, Issac, Lee, Soo, and Kim, Woohyun

Subjects
Animals, Gap Junctions, Interneurons, Male, Mice, Mice, Knockout, Nerve Tissue Proteins, Parvalbumins, Receptors, N-Methyl-D-Aspartate, Social Behavior, Social Cognition, Synapses
Abstract

NMDA receptor (NMDAR) and GABA neuronal dysfunctions are observed in animal models of autism spectrum disorders, but how these dysfunctions impair social cognition and behavior remains unclear. We report here that NMDARs in cortical parvalbumin (Pv)-positive interneurons cooperate with gap junctions to promote high-frequency (>80 Hz) Pv neuronal burst firing and social cognition. Shank2-/- mice, displaying improved sociability upon NMDAR activation, show impaired cortical social representation and inhibitory neuronal burst firing. Cortical Shank2-/- Pv neurons show decreased NMDAR activity, which suppresses the cooperation between NMDARs and gap junctions (GJs) for normal burst firing. Shank2-/- Pv neurons show compensatory increases in GJ activity that are not sufficient for social rescue. However, optogenetic boosting of Pv neuronal bursts, requiring GJs, rescues cortical social cognition in Shank2-/- mice, similar to the NMDAR-dependent social rescue. Therefore, NMDARs and gap junctions cooperate to promote cortical Pv neuronal bursts and social cognition.

Published
2021

10. Intra-class Contrastive Learning Improves Computer Aided Diagnosis of Breast Cancer in Mammography

Author

You, Kihyun, Lee, Suho, Jo, Kyuhee, Park, Eunkyung, Kooi, Thijs, Nam, Hyeonseob, Goos, Gerhard, Founding Editor, Hartmanis, Juris, Founding Editor, Bertino, Elisa, Editorial Board Member, Gao, Wen, Editorial Board Member, Steffen, Bernhard, Editorial Board Member, Yung, Moti, Editorial Board Member, Wang, Linwei, editor, Dou, Qi, editor, Fletcher, P. Thomas, editor, Speidel, Stefanie, editor, and Li, Shuo, editor

Published
2022
Full Text
View/download PDF

16. Loss of Katnal2 leads to ependymal ciliary hyperfunction and autism-related phenotypes in mice

Author

Kang, Ryeonghwa, primary, Kim, Kyungdeok, additional, Jung, Yewon, additional, Choi, Sang-Han, additional, Lee, Chanhee, additional, Im, Geun Ho, additional, Shin, Miram, additional, Ryu, Kwangmin, additional, Choi, Subin, additional, Yang, Esther, additional, Shin, Wangyong, additional, Lee, Seungjoon, additional, Lee, Suho, additional, Papadopoulos, Zachary, additional, Ahn, Ji Hoon, additional, Koh, Gou-Young, additional, Kipnis, Jonathan, additional, Kang, Hyojin, additional, Kim, Hyun, additional, Cho, Won-Ki, additional, Park, Soochul, additional, Kim, Seong-Gi, additional, and Kim, Eunjoon, additional

Published
2024
Full Text
View/download PDF

22. Rescue of Transgenic Alzheimer’s Pathophysiology by Polymeric Cellular Prion Protein Antagonists

Author

Gunther, Erik C., Smith, Levi M., Kostylev, Mikhail A., Cox, Timothy O., Kaufman, Adam C., Lee, Suho, Folta-Stogniew, Ewa, Maynard, George D., Um, Ji Won, Stagi, Massimiliano, Heiss, Jacqueline K., Stoner, Austin, Noble, Geoff P., Takahashi, Hideyuki, Haas, Laura T., Schneekloth, John S., Merkel, Janie, Teran, Christopher, Naderi, Zahra K., Supattapone, Surachai, and Strittmatter, Stephen M.

Published
2019
Full Text
View/download PDF

24. SLC6A20 transporter: a novel regulator of brain glycine homeostasis and NMDAR function

Author

Bae, Mihyun, Roh, Junyeop Daniel, Kim, Youjoung, Kim, Seong Soon, Han, Hye Min, Yang, Esther, Kang, Hyojin, Lee, Suho, Kim, Jin Yong, Kang, Ryeonghwa, Jung, Hwajin, Yoo, Taesun, Kim, Hyosang, Kim, Doyoun, Oh, Heejeong, Han, Sungwook, Kim, Dayeon, Han, Jinju, Bae, Yong Chul, Kim, Hyun, Ahn, Sunjoo, Chan, Andrew M, Lee, Daeyoup, Kim, Jin Woo, and Kim, Eunjoon

Published
2021
Full Text
View/download PDF

30. Unbounded Scalable Verification Based on Approximate Property-Directed Reachability and Datapath Abstraction

Author

Lee, Suho, Sakallah, Karem A., Hutchison, David, editor, Kanade, Takeo, editor, Kittler, Josef, editor, Kleinberg, Jon M., editor, Kobsa, Alfred, editor, Mattern, Friedemann, editor, Mitchell, John C., editor, Naor, Moni, editor, Nierstrasz, Oscar, editor, Pandu Rangan, C., editor, Steffen, Bernhard, editor, Terzopoulos, Demetri, editor, Tygar, Doug, editor, Weikum, Gerhard, editor, Biere, Armin, editor, and Bloem, Roderick, editor

Published
2014
Full Text
View/download PDF

37. IRSp53 Promotes Postsynaptic Density Formation and Actin Filament Bundling

Author

Feng, Zhe, Lee, Suho, Jia, Bowen, Jian, Tao, Kim, Eunjoon, Zhang, Mingjie, Feng, Zhe, Lee, Suho, Jia, Bowen, Jian, Tao, Kim, Eunjoon, and Zhang, Mingjie

Abstract

IRSp53 (aka BAIAP2) is a scaffold protein that couples membranes with the cytoskeleton in actin-filled protrusions such as filopodia and lamellipodia. The protein is abundantly expressed in excitatory synapses and is essential for synapse development and synaptic plasticity, although with poorly understood mechanisms. Here we show that specific multivalent interactions between IRSp53 and its binding partners PSD-95 or Shank3 drive phase separation of the complexes in solution. IRSp53 can be enriched to the reconstituted excitatory PSD (ePSD) condensates via bridging to the core and deeper layers of ePSD. Overexpression of a mutant defective in the IRSp53/PSD-95 interaction perturbs synaptic enrichment of IRSp53 in mouse cortical neurons. The reconstituted PSD condensates promote bundled actin filament formation both in solution and on membranes, via IRSp53-mediated actin binding and bundling. Overexpression of mutants that perturb IRSp53–actin interaction leads to defects in synaptic maturation of cortical neurons. Together, our studies provide potential mechanistic insights into the physiological roles of IRSp53 in synapse formation and function.

Published
2022

40. Kv7/KCNQ potassium channels in cortical hyperexcitability and juvenile seizure-related death in Ank2-mutant mice.

Author

Oh, Hyoseon, Lee, Suho, Oh, Yusang, Kim, Seongbin, Kim, Young Seo, Yang, Yeji, Choi, Woochul, Yoo, Ye-Eun, Cho, Heejin, Lee, Seungjoon, Yang, Esther, Koh, Wuhyun, Won, Woojin, Kim, Ryunhee, Lee, C. Justin, Kim, Hyun, Kang, Hyojin, Kim, Jin Young, Ku, Taeyun, and Paik, Se-Bum

Subjects
POTASSIUM channels, AUTISM spectrum disorders, SCAFFOLD proteins, MICE
Abstract

Autism spectrum disorders (ASD) represent neurodevelopmental disorders characterized by social deficits, repetitive behaviors, and various comorbidities, including epilepsy. ANK2, which encodes a neuronal scaffolding protein, is frequently mutated in ASD, but its in vivo functions and disease-related mechanisms are largely unknown. Here, we report that mice with Ank2 knockout restricted to cortical and hippocampal excitatory neurons (Ank2-cKO mice) show ASD-related behavioral abnormalities and juvenile seizure-related death. Ank2-cKO cortical neurons show abnormally increased excitability and firing rate. These changes accompanied decreases in the total level and function of the Kv7.2/KCNQ2 and Kv7.3/KCNQ3 potassium channels and the density of these channels in the enlengthened axon initial segment. Importantly, the Kv7 agonist, retigabine, rescued neuronal excitability, juvenile seizure-related death, and hyperactivity in Ank2-cKO mice. These results suggest that Ank2 regulates neuronal excitability by regulating the length of and Kv7 density in the AIS and that Kv7 channelopathy is involved in Ank2-related brain dysfunctions. Seizures can occur in autism spectrum disorders (ASD). Here, the authors report that mice lacking the ASD related Ank2 gene show abnormal behaviours, and juvenile seizure related death, through impaired Kv7 potassium channel function. [ABSTRACT FROM AUTHOR]

Published
2023
Full Text
View/download PDF

45. Image-Based Analysis of the Genome’s Fractality During the Cell Cycle

Author

Lee, Suho, Liu, Xutong, Ziabkin, Ivan, and Zidovska, Alexandra

Abstract

The human genome consists of about 2 m of DNA packed inside the cell nucleus barely 10 μm in diameter. DNA is complexed with histones, forming chromatin fiber, which folds inside the nucleus into loops, TADs, A/B compartments and chromosome territories. This organization is knot-free and self-similar across length scales, leading to a hypothesis that the genome presents a fractal globule, which was corroborated by chromosome conformation capture experiments. In addition, many microscopy techniques have been used to obtain the fractal dimension of the genome’s spatial distribution from its images. However, different techniques often required that different definitions of fractal dimension be adapted, making the comparison of these results not trivial. In this study, we use spinning disc confocal microscopy to collect high-resolution images of nuclei in live human cells during the cell cycle. We then systematically compare existing image-based fractal analyses – including mass-scaling, box-counting, lacunarity and multifractal spectrum – by applying them to images of human cell nuclei and investigate changes in the genome’s spatial organization during the cell cycle. Our data reveal that different image-based fractal measurements offer distinct metrics, highlighting different features of the genome’s spatial organization. Yet, all these metrics consistently indicate the following trend for the changes in the genome’s organization during the cell cycle: the genome being compactly packed in early G1 phase, followed by a decondensation throughout the G1 phase, and a subsequent condensation in the S and G2 phases. Our comprehensive comparison of image-based fractal analyses reconciles the perceived discrepancies between different methods. Moreover, our results offer new insights into the physical principles underlying the genome’s organization and its changes during the cell cycle.

Published
2025
Full Text
View/download PDF

50. A Low-Power 65/14nm Stacked CMOS Image Sensor

Author

Kwon, Minho, primary, Lim, Seunghyun, additional, Lee, Hyeokjong, additional, Ha, Il-Seon, additional, Kim, Moo-Young, additional, Seo, Il-Jin, additional, Lee, Suho, additional, Choi, Yongsuk, additional, Kim, Kyunghoon, additional, Lee, Hansoo, additional, Kim, Won-Woong, additional, Park, Seonghye, additional, Koh, Kyongmin, additional, Lee, Jesuk, additional, and Park, Yongin, additional

Published
2020
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results