Korea Advanced Institute of Science and Technology · Physics and Astronomy
Professor Jihoon Ko's research lab specializes in machine learning and deep learning applications for structured and sequential data, with a strong focus on graph-structured data, temporal modeling, and efficient representation learning. The lab explores continual learning for graphs, influence maximization in social networks, and end-to-end learning for tensor compression and weather prediction. Key research directions include developing scalable, generalizable, and efficient deep learning models that handle real-world complexities such as data sparsity, distribution shift, and long-term dependencies. The lab emphasizes practical deployment through lightweight, high-performance methods that balance accuracy and computational efficiency.
Figures are computed from collected data and may differ slightly.
Hypergraphs naturally represent group interactions, which are omnipresent in many domains: collaborations of researchers, co-purchases of items, joint interactions of proteins, to name a few. In this work, we propose tools for answering the following questions in a systematic manner: ( Q1 ) what are structural design principles of real-world hypergraphs? ( Q2 ) how can we compare local structures of hypergraphs of different sizes? ( Q3 ) how can we identify domains which hypergraphs are from? We
Angiogenic sprouting, which is the growth of new blood vessels from pre-existing vessels, is orchestrated by cues from the cellular microenvironment, such as spatially controlled gradients of angiogenic factors. However, current in vitro models are less scalable for in-depth studies of angiogenesis. In this study, a plastic-based microfluidic chip is developed to reconstruct in vitro 3D vascular networks. The main disadvantages of the preexisting system are identified, namely, the low productivi
We guide the use of organ-on-chip technology in tissue engineering applications. Organ-on-chip technology is a form of microengineered cell culture platform that elaborates the in-vivo like organ or tissue microenvironments. The organ-on-chip platform consists of microfluidic channels, cell culture chambers, and stimulus sources that emulate the in-vivo microenvironment. These platforms are typically engraved into an oxygen-permeable transparent material. Fabrication of these materials requires
Gastric cancer (GC) with peritoneal metastases and malignant ascites continues to have poor prognosis. Exosomes mediate intercellular communication during cancer progression and promote therapeutic resistance. Here, we report the significance of exosomes derived from malignant ascites (EXO<sup>Ascites</sup>) in cancer progression and use modified exosomes as resources for cancer therapy. EXO<sup>Ascites</sup> from patients with GC stimulated invasiveness and angiogenesis in an ex vivo three-dime
What kind of macroscopic structural and dynamical patterns can we observe in real-world hypergraphs? What can be underlying local dynamics on individuals, which ultimately lead to the observed patterns, beyond apparently random evolution? Graphs, which provide effective ways to represent pairwise interactions among entities, fail to represent group interactions (e.g., collaboration of three or more researchers, etc.). Regarded as a generalization of graphs, hypergraphs allowing for various sizes
Accurate nowcasting is critical for preemptive action in response to heavy rainfall events (HREs). However, operational numerical weather prediction models have difficulty predicting HREs in the short term, especially for rapidly and sporadically developing cases. Here, we present multi-year evaluation statistics showing that deep-learning-based HRE nowcasting, trained with radar images and ground measurements, outperforms short-term numerical weather prediction at lead times of up to 6 h. The d
Cancer metastasis, the leading cause of cancer-related deaths, remains a complex challenge in medical science. Stephen Paget's "seed and soil theory" introduced the concept of organotropism, suggesting that metastatic success depends on specific organ microenvironments. Understanding organotropism not only offers potential for curbing metastasis but also novel treatment strategies. Microphysiological systems (MPS), especially organ-on-a-chip models, have emerged as transformative tools in this q
Given a massive graph, how can we exploit its hierarchical structure for concisely but exactly summarizing the graph? By exploiting the structure, can we achieve better compression rates than state-of-the-art graph summarization methods? The explosive proliferation of the Web has accelerated the emergence of large graphs, such as online social networks and hyperlink networks. Consequently, graph compression has become increasingly important to process such large graphs without expensive I/O over
<b>Background:</b> To aid in oncology drug development, we investigated MET proto-oncogene receptor tyrosine kinase gene aberrations in 2,239 oncology patients who underwent next-generation sequencing (NGS) in clinical practice. <b>Materials and methods:</b> From November 2019 to January 2021, 2,239 patientswith advanced solid tumors who visited oncology clinics underwent NGS. The NGS panel included >500 comprehensive NGS tests using archival tissue specimens. Programmed death-ligand 1(PD-L1) 22
Open papers in the app to read, cite, and organize with AI.
