Colloquia and Seminars

Selected for an oral presentation at CPRV 2026; Image classification is a well-studied task in computer vision, and yet it remains challenging under high-uncertainty conditions, such as when input images are corrupted or training data are limited. Conventional classification approaches typically train models to directly predict class labels from input images, but this might lead to suboptimal performance in such scenarios. To address this issue, we propose Discrete Diffusion Classification Modeling (DiDiCM), a novel framework that leverages a diffusion-based procedure to model the posterior distribution of class labels conditioned on the input image. DiDiCM supports diffusion-based predictions either on class probabilities or on discrete class labels, providing flexibility in computation and memory trade-offs. We conduct a comprehensive empirical study demonstrating the superior performance of DiDiCM over standard classifiers, showing that a few diffusion iterations achieve higher classification accuracy on the ImageNet dataset compared to baselines, with accuracy gains increasing as the task becomes more challenging.

High-quality multimodal neuroimaging provides complementary information essential to both neuroscience and neurology. However, due to multifaceted practical limitations, complete imaging data is often unavailable in scenarios such as emergency treatment or routine screening. Generative models offer a promising paradigm to address these bottlenecks by imputing missing modalities, enhancing image quality,  harmonizing acquisition domain discrepancies, and simulating diverse disease-relevant appearances. This presentation will explore algorithmic innovations in generative AI, including a variety of model architectures and training strategies, with a focus on their versatile applications across multimodal image generation, cross-field MRI synthesis, image super-resolution, and disease-specific image simulation. Furthermore, we will demonstrate how these synthesized and enhanced images effectively facilitate critical downstream tasks, such as subcortical segmentation, cross-modal registration, and clinical diagnosis, with the ultimate goal of improving patient outcomes.

Yulin Wang is a Postdoctoral Researcher at the School of Biomedical Engineering, ShanghaiTech University. In September 2026, she will transition to a postdoctoral position at Cornell Tech, the joint academic venture of Cornell University and the Technion. Her research interests lie at the intersection of medical image computing and analysis, generative artificial intelligence, and neuroradiology, with a primary focus on developing advanced deep learning frameworks to solve challenging reconstruction, synthesis, and enhancement problems in clinical neuroimaging. Her recent works have been published in top-tier journals and conferences, including Cell Reports Medicine, Medical Physics, Physics in Medicine and Biology, and MICCAI.

In the Degree Realization problem with respect to a family P of graphs the input is a non-increasing sequence d = (d1, . . . , dn) of positive integers, and the goal is to decide whether there exists a simple undirected graph G ∈ P, whose degrees correspond to d, i.e., such that deg(G) = d. In this paper we consider the version of Degree Realization in which the realization is required to be a forest (i.e., P is the family for forests). We consider optimized Degree Realization in which the goal is to obtain a realization that minimizes an objective function f. That is, the goal is to find a realization G that minimizes f(G) among the realizations of the given input sequence. More specifically, we focus on the following functions: the size of an optimal vertex cover and the size of an optimal dominating set. We also consider the total and paired versions of both Min Vertex Cover and Min Dominating Set. We provide characterizations and linear time realization algorithms for all the above-mentioned problems.