BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Biomedical Mathematics Group - ECPv6.16.2//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Biomedical Mathematics Group
X-ORIGINAL-URL:https://www.ibs.re.kr/bimag
X-WR-CALDESC:Events for Biomedical Mathematics Group
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:Asia/Seoul
BEGIN:STANDARD
TZOFFSETFROM:+0900
TZOFFSETTO:+0900
TZNAME:KST
DTSTART:20240101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=Asia/Seoul:20250703T160000
DTEND;TZID=Asia/Seoul:20250703T170000
DTSTAMP:20260521T160346
CREATED:20250628T074404Z
LAST-MODIFIED:20250630T055202Z
UID:11207-1751558400-1751562000@www.ibs.re.kr
SUMMARY:Jihun Han - Bridging PDEs and machine learning
DESCRIPTION:Abstract: This talk consists of two main parts. In the first part\, I will discuss a numerical method for solving PDEs based on a stochastic representation of the solution. This approach captures the underlying particle dynamics associated with the physical processes described by the PDE. By aggregating information from the particles’ collective exploration\, the method iteratively reinforces the approximation toward the solution. I will cover its analysis regarding the trainability and highlight its effectiveness across a broad class of problems\, including elliptic equations with interfaces\, multiscale structures\, and perforated domains\, as well as hyperbolic-type problems such as the Eikonal and Burgers equations.\nIn the second part\, I will present a method for learning in-between imagery dynamics. This approach integrates PDE models within latent spaces to enhance both learning capability and interpretability. Notably\, this method demonstrates robustness in capturing intricate dynamics\, such as rotation and outflow\, which pose significant challenges for current state-of-the-art optimal transport methods.
URL:https://www.ibs.re.kr/bimag/event/ji-hoon-han-bridging-pdes-and-machine-learning/
LOCATION:B232 Seminar Room\, IBS\, 55 Expo-ro Yuseong-gu\, Daejeon\, Daejeon\, 34126\, Korea\, Republic of
CATEGORIES:Biomedical Mathematics Seminar
ORGANIZER;CN="Jae Kyoung Kim":MAILTO:jaekkim@kaist.ac.kr
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Seoul:20250704T140000
DTEND;TZID=Asia/Seoul:20250704T160000
DTSTAMP:20260521T160346
CREATED:20250526T004910Z
LAST-MODIFIED:20250609T001902Z
UID:11146-1751637600-1751644800@www.ibs.re.kr
SUMMARY:Machine learning methods trained on simple models can predict critical transitions in complex natural systems - Shingo Gibo
DESCRIPTION:In this talk\, we discuss the paper “Machine learning methods trained on simple models can predict critical transitions in complex natural systems” by  Smita Deb\, Sahil Sidheekh\, Christopher F. Clements\, Narayanan C. Krishnan\, and Partha S. Dutta\, in Royal Society Open Science\, (2022). \nAbstract:  \nForecasting sudden changes in complex systems is a critical but challenging task\, with previously developed methods varying widely in their reliability. Here we develop a novel detection method\, using simple theoretical models to train a deep neural network to detect critical transitions—the Early Warning Signal Network (EWSNet). We then demonstrate that this network\, trained on simulated data\, can reliably predict observed real-world transitions in systems ranging from rapid climatic change to the collapse of ecological populations. Importantly\, our model appears to capture latent properties in time series missed by previous warning signals approaches\, allowing us to not only detect if a transition is approaching\, but critically whether the collapse will be catastrophic or non-catastrophic. These novel properties mean EWSNet has the potential to serve as an indicator of transitions across a broad spectrum of complex systems\, without requiring information on the structure of the system being monitored. Our work highlights the practicality of deep learning for addressing further questions pertaining to ecosystem collapse and has much broader management implications.
URL:https://www.ibs.re.kr/bimag/event/machine-learning-methods-trained-on-simple-models-can-predict-critical-transitions-in-complex-natural-systems-shingo-gibo/
LOCATION:B232 Seminar Room\, IBS\, 55 Expo-ro Yuseong-gu\, Daejeon\, Daejeon\, 34126\, Korea\, Republic of
CATEGORIES:Journal Club
ORGANIZER;CN="Jae Kyoung Kim":MAILTO:jaekkim@kaist.ac.kr
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Seoul:20250711T140000
DTEND;TZID=Asia/Seoul:20250711T160000
DTSTAMP:20260521T160346
CREATED:20250628T122808Z
LAST-MODIFIED:20250628T122808Z
UID:11216-1752242400-1752249600@www.ibs.re.kr
SUMMARY:Optimal transport for generating transition states in chemical reactions - Gyuyoung Hwang
DESCRIPTION:In this talk\, we discuss the paper “Optimal transport for generating transition states in chemical reactions” by C. Duan et.al.\, Nat. Machine. Intelligence\, 2025. \nAbstract \nTransition states (TSs) are transient structures that are key to understanding reaction mechanisms and designing catalysts but challenging to capture in experiments. Many optimization algorithms have been developed to search for TSs computationally. Yet\, the cost of these algorithms driven by quantum chemistry methods (usually density functional theory) is still high\, posing challenges for their applications in building large reaction networks for reaction exploration. Here we developed React-OT\, an optimal transport approach for generating unique TS structures from reactants and products. React-OT generates highly accurate TS structures with a median structural root mean square deviation of 0.053 Å and median barrier height error of 1.06 kcal mol−1 requiring only 0.4 s per reaction. The root mean square deviation and barrier height error are further improved by roughly 25% through pretraining React-OT on a large reaction dataset obtained with a lower level of theory\, GFN2-xTB. We envision that the remarkable accuracy and rapid inference of React-OT will be highly useful when integrated with the current high-throughput TS search workflow. This integration will facilitate the exploration of chemical reactions with unknown mechanisms.
URL:https://www.ibs.re.kr/bimag/event/optimal-transport-for-generating-transition-states-in-chemical-reactions-gyuyoung-hwang/
LOCATION:B232 Seminar Room\, IBS\, 55 Expo-ro Yuseong-gu\, Daejeon\, Daejeon\, 34126\, Korea\, Republic of
CATEGORIES:Journal Club
ORGANIZER;CN="Jae Kyoung Kim":MAILTO:jaekkim@kaist.ac.kr
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Seoul:20250718T140000
DTEND;TZID=Asia/Seoul:20250718T160000
DTSTAMP:20260521T160346
CREATED:20250701T022224Z
LAST-MODIFIED:20250701T022224Z
UID:11231-1752847200-1752854400@www.ibs.re.kr
SUMMARY:scGPT: toward building a foundation model for single-cell multi-omics using generative AI - Hyun Kim
DESCRIPTION:In this talk\, we discuss the paper “scGPT: toward building a foundation model for single-cell multi-omics using generative AI” by Haotian Cui\, et.al. Nature Methods\, 2024. \nAbstract \nGenerative pretrained models have achieved remarkable success in various domains such as language and computer vision. Specifically\, the combination of large-scale diverse datasets and pretrained transformers has emerged as a promising approach for developing foundation models. Drawing parallels between language and cellular biology (in which texts comprise words; similarly\, cells are defined by genes)\, our study probes the applicability of foundation models to advance cellular biology and genetic research. Using burgeoning single-cell sequencing data\, we have constructed a foundation model for single-cell biology\, scGPT\, based on a generative pretrained transformer across a repository of over 33 million cells. Our findings illustrate that scGPT effectively distills critical biological insights concerning genes and cells. Through further adaptation of transfer learning\, scGPT can be optimized to achieve superior performance across diverse downstream applications. This includes tasks such as cell type annotation\, multi-batch integration\, multi-omic integration\, perturbation response prediction and gene network inference.
URL:https://www.ibs.re.kr/bimag/event/scgpt-toward-building-a-foundation-model-for-single-cell-multi-omics-using-generative-ai-hyun-kim/
LOCATION:B232 Seminar Room\, IBS\, 55 Expo-ro Yuseong-gu\, Daejeon\, Daejeon\, 34126\, Korea\, Republic of
CATEGORIES:Journal Club
ORGANIZER;CN="Jae Kyoung Kim":MAILTO:jaekkim@kaist.ac.kr
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Seoul:20250721T110000
DTEND;TZID=Asia/Seoul:20250721T120000
DTSTAMP:20260521T160346
CREATED:20250617T084231Z
LAST-MODIFIED:20250617T084231Z
UID:11189-1753095600-1753099200@www.ibs.re.kr
SUMMARY:Jae-Kwang Kim - Weight calibration for causal inference and transfer learning
DESCRIPTION:Abstract: Weight calibration is a popular technique in handling covariate-shift problem in causal inference. It can be viewed as a dual optimization problem for incorporating the implicit regression model. We introduce the generalized entropy calibration as a general tool for weight calibration. Several interesting applications will be introduced in the context of causal inference. Furthermore\, weight calibration can be used to transfer learning\, which combines information from two different samples\, one for source data and the other for target data.
URL:https://www.ibs.re.kr/bimag/event/jae-kwang-kim-weight-calibration-for-causal-inference-and-transfer-learning/
LOCATION:B232 Seminar Room\, IBS\, 55 Expo-ro Yuseong-gu\, Daejeon\, Daejeon\, 34126\, Korea\, Republic of
CATEGORIES:Biomedical Mathematics Seminar
ORGANIZER;CN="Jae Kyoung Kim":MAILTO:jaekkim@kaist.ac.kr
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=Asia/Seoul:20250725T140000
DTEND;TZID=Asia/Seoul:20250725T160000
DTSTAMP:20260521T160346
CREATED:20250628T123019Z
LAST-MODIFIED:20250721T002532Z
UID:11218-1753452000-1753459200@www.ibs.re.kr
SUMMARY:Effective Markovian dynamics method of solving non-Markovian dynamics of stochastic gene expression - Dongju Lim
DESCRIPTION:In this talk\, we discuss the paper “Effective Markovian dynamics method of solving non-Markovian dynamics of stochastic gene expression” by Youming Li and Chen Jia\, Physical Review Letters\, to appear. \nAbstract \nExperiments have shown that over 10% of proteins are degraded non-exponentially. Gene expression models for non-exponentially degraded proteins are notoriously difficult to solve since the underlying stochastic dynamics is non-Markovian. Here we develop an effective Markovian dynamics (EMD) method which converts a large class of non-Markovian models into effective Markovian ones so that they have the same mRNA and protein distributions at any fixed time. Using the EMD approach\, we analytically solve some classical gene expression models with non-exponential or delayed protein decay\, whose exact distributions are previously unknown and fail to be obtained using conventional queueing theory. Our theory successfully explains why non-exponentially degraded proteins on average have smaller mRNA-protein correlation than exponentially degraded proteins\, and it predicts that bimodality is significantly enhanced in the presence of delayed protein degradation.
URL:https://www.ibs.re.kr/bimag/event/action-functional-as-an-early-warning-indicator-in-the-space-of-probability-measures-via-schrodinger-bridge-dongju-lim/
LOCATION:B232 Seminar Room\, IBS\, 55 Expo-ro Yuseong-gu\, Daejeon\, Daejeon\, 34126\, Korea\, Republic of
CATEGORIES:Journal Club
ORGANIZER;CN="Jae Kyoung Kim":MAILTO:jaekkim@kaist.ac.kr
END:VEVENT
END:VCALENDAR