Speaker: Paul A. YushkevichVenue: Room 414, CHOW YEI CHING BUILDING, Zijingang CampusAbstract: Dr. Paul Yushkevich is a Professor in the Department of Radiology at the University of Pennsylvania and a member of the Bioengineering Graduate Group. His research focuses on developing novel computational methodologies for biomedical image analysis, including shape representation for statistical analysis, multi-atlas and shape-based segmentation, groupwise registration, and structure-specific fMRI and diffusion MRI analysis. In his graduate work at UNC under Stephen M. Pizer, he developed the continuous medial representation (cm-rep) approach, and he has since extended this work using differential equations to solve complex geometric constraints in skeleton-based shape analysis. He has built a detailed atlas of the hippocampal region from high-field, ultra-high-resolution MRI and dense histology, and has applied his algorithms to problems in neuroimaging and cardiac imaging – earning first-place finishes in MICCAI segmentation challenges in 2012 and 2013. Dr. Yushkevich is also deeply committed to open-source software, leading the development of ITK-SNAP (a widely used interactive 3D segmentation tool with >10,000 monthly downloads) and the companion Convert3D utility. His recent work includes ex vivo 3D mapping of tau pathology in the medial temporal lobe, the ASHS software for hippocampal subfield segmentation, and deep learning-based biomarkers for early Alzheimer's disease progression.

Speaker: Bak Kong LowVenue: Room 414, Haigong Building, Zhoushan CampusAbstract: This talk presents examples of efficient first- and second-order reliability methods (FORM and SORM) in environmental and water engineering. It will be shown that system reliability, importance sampling, and Monte Carlo simulations can be done as lucid extensions of FORM. The aim is to overcome conceptual and implementation hurdles of probability methods for engineers, advanced undergraduates, and master and PhD students. Instead of lengthy mathematics in the traditional approach, the same solutions are obtained quickly and clearly by automatic cell-based constrained optimization in a ubiquitous platform. The talk will focus on insights, alternative perspectives and comparisons with other methods.

Speaker: Khellil SefianeVenue: Room 2-328, School of Mechanical Engineering Building, Zijingang CampusAbstract: The lecture will present the results of a series of studies on droplets evaporation and wetting. It will aim to elucidate the effects of substrate thermal properties as well as atmosphere nature and atmospheric conditions. The presentation will introduce the results also reveal some hydrothermal instabilities and patterns. The evaporation of complex fluids drops such as nanofluids and polymers will be introduced. Theoretical studies on evaporating droplets lifetimes and effect of wettability will be discussed. Following these fundamental aspects of droplets evaporation the lecture will introduce some concepts on the development of a Leidenfrost engine driven by droplets levitating on hot substrates.

Speaker: HAN YanVenue: Room A423, School of Management, Zijingang CampusAbstract: Han Yan is an assistant professor of accounting at the Sauder School of Business, University of British Columbia. His research focuses on banking, climate risk, and artificial intelligence. He received his PhD in Accounting from the Stern School of Business at New York University. His recent work has been published in the Review of Accounting Studies.

Speaker: Ray ZhangVenue: Room A423, School of Management, Zijingang CampusAbstract: Using a proprietary dataset that documents corporate bond issuances in China, we find that bond regulators' comment letters (CLs) on issuers' public disclosures generally target opaque firms and reduce bond issuance success. We also find, however, that connected investors—financial institutions with prior business relationships with the issuers—purchase a relatively larger share of bond issuances from CL recipients than from non-CL recipients, and this greater participation contributes to overpricing. Further analysis shows that when connected investors support CL recipients, they are motivated by their capacity to transfer private information and benefits derived from their connections.

Speaker: Christian OliversVenue: Room 313, Haina Building 3, Zijingang CampusAbstract: Chris Olivers (MSc Nijmegen, Netherlands, 1996; PhD Birmingham, UK 2001) is Full Professor of Visual Cognition, principal investigator at the Institute for Brain and Behaviour Amsterdam, and Head of Department of Experimental and Applied Psychology at the Vrije Universiteit. His main research lines focus on attention, working memory and cognitive control, using behavioral and neurophysiological approaches. His early work has been recognized with early career awards from the APA and British Psychological Society. His later work has been funded by prestigious NWO Veni, Vidi, Vici, and ERC Consolidator grants, with high profile publications in journals like Psychological Review, PNAS, Journal of Neuroscience, Cerebral Cortex, Psychological Science, and Trends in Cognitive Sciences. More recently he also moved into more applied fields of vision impairment and safety design. He has coordinated several international research networks and has been Chief Editor of Visual Cognition. He also served as the chair and co-founder of the National Ethics Council in Social and Behavioural Sciences in the Netherlands (nethics.nl), and is currently a government-appointed expert member of the Central Committee for Human Research in the Netherlands.Visual working memory (VWM) refers to the cognitive mechanisms that allow us to temporarily retain task relevant visual information. Traditionally, VWM has been studied as a memory of something, emphasizing the capacity and fidelity of maintained representations. More recent work, however, frames VWM as a memory for something—highlighting its prospective role in guiding goals, attention, and action. In this talk, I will review several lines of research from my lab that investigate VWM as a function of the upcoming task. These include memory for attention, memory for action, and memory for task scheduling. Across these projects, we show how the prospective purpose of VWM reshapes its representations and the control processes that govern them.

Speaker: Andrei OkounkovVenue: Room 210, Haina Building 2, Zijingang CampusAbstract: Enumerative geometers and their colleagues in representation theory and mathematical physics are very excited about the new perspectives on old and new problems offered by the nascent field of 3-dimensional mirror symmetry. While most formulations or explanations of what 3-dimensional mirror symmetry is require a lot of prerequisites and a high level of abstraction, some of its core predictions can be easily cast in the language that people like P.L. Chebyshev, C.G. Jacobi, and I.G. Macdonald would have no problem grasping. This is what I will try to do in this talk, which I hope will be accessible to the general mathematical audience.

Speaker: Brian TomVenue: Room 210, Haina Building 2, Zijingang CampusAbstract: In precision medicine, there is growing interest in discovering disease endotypes from high-dimensional biomarkers which associate with clinically relevant outcomes or phenotypes. Motivated by an application in knee osteoarthritis (OA) using synovial fluid protein markers, we aimed to identify OA endotypes linked to disease severity (based on Kellgren–Lawrence grade). Bayesian profile regression (Molitor et al. 2010) was adopted to perform this outcome-guided clustering. However, although the clusters found were stable across training and validation sets, the prediction of disease severity from these clusters were poor in the validation set; thus, limiting its practical use. The poor out-of-sample performance can partly be attributed to the weak influence of the single binary outcome relative to the high dimensional set of biomarkers in determining the clustering structure. To address this imbalance and for more general multi-task learning problems, we propose a generalized Bayesian framework for outcome-guided clustering by reformulating the problem as a decision problem with separate loss components for the clustering and prediction tasks. This framework offers greater flexibility and admits standard Bayesian profile regression as a special case. To improve the interpretability of the weighting of the different loss components, we propose a principled standardization that places each task risk on a common scale, thereby decoupling the task weights from the learning rate. The learning rate is tuned/guided by an estimate of the expected out-of-sample risk under a user-chosen evaluation loss.

Speaker: Joy TongVenue: Room 530, Chengjun Building 7, Zijingang CampusAbstract: We use a novel text-based classification method to identify green trademarks that capture environmentally friendly products and services introduced to the market. Firms with more green products generate higher green revenues, emit less greenhouse gas, and receive higher environmental ratings. These firms also exhibit higher firm value, particularly when green products align with core businesses and prior experience and generate product synergies. The valuation effect appears to be driven by consumer environmental awareness rather than investor preferences or concerns. Finally, we provide causal evidence that firms expand their green product offerings following nearby natural disasters or peer firms’ environmental scandals.