Siqi Wang

I am a 5th year Computer Science PhD student at Boston University, Image and Video Computing Group, where I'm fortunate to be advised by Prof. Bryan A. Plummer. Before joining BU, I was advised by Prof. Daniel Ritchie at Brown Visual Computing and received my Master degree in Computer Science in 2020. I received my Bachelor’s degree in Computer Science at Beihang University, Beijing, China in 2018.

siqiwang@bu.edu / CV / Google Scholar / Github / LinkedIn

Teaching Fellow

Research

My research interests lie within machine learning and computer vision, with a focus on learning with noisy labels (LNL) and multi-distribution data (Domain Generalization). Specifically, I work on improving noise detection and enhancing generalization performance. My research has applications in biomedical data, where I aim to advance machine intelligence for medical insights such as drug discovery. And in e-commerce, where I focus on designing search algorithms that handle noisy and multi-regional data effectively.

	Noise-Aware Generalization: Robustness to In-Domain Noise and Out-of-Domain Generalization Siqi Wang, Aoming Liu, Bryan A. Plummer, arXiv, 2025 arXiv We investigate the underexplored space in Domain Generalization (DG), where models are evaluated under both distribution shifts and label noise, which we refer to as Noise-Aware Generalization (NAG). In NAG, distribution shifts can be due to label noise or domain shifts, breaking the assumptions used by Learning with Noisy Labels (LNL) methods. Our proposed DL4ND approach improves noise detection by taking advantage of the observation that noisy samples that may appear indistinguishable within a single domain often show greater variation when compared across domains.
	SEQ+MD:Learning Multi-Task as a SEQuence with MultiDistribution Data Siqi Wang, Audrey Zhijiao Chen, Austin Clapp, Sheng-Min Shih, Xiaoting Zhao, Accepted by ACM SIGIR Workshop on eCommerce, 2025 arXiv In this work, we propose the SEQ+MD framework, which integrates sequential learning for multi-task learning (MTL) and feature-generated region-mask for multi-distribution input. This approach leverages the sequential order within tasks and accounts for regional heterogeneity, enhancing performance on multi-source data.
	LNL+K: Learning with Noisy Labels and Noise Source Distribution Knowledge Siqi Wang, Bryan A. Plummer, The European Conference on Computer Vision (ECCV), 2024 arXiv We introduce a new task called Learning with Noisy Labels and noise source distribution Knowledge (LNL+K), which assumes we have some knowledge about likely source(s) of label noise that we can take advantage of. By making this presumption, methods are better equipped to distinguish hard negatives between categories from label noise. In addition, this enables us to explore datasets where the noise may represent the majority of samples, a setting that breaks a critical premise of most methods developed for the LNL task.
	A Unified Framework for Connecting Noise Modeling to Boost Noise Detection Siqi Wang, Chau Pham, Bryan A. Plummer, arXiv, 2023 arXiv In this work, we explore the integration of noise modeling and noise detection, proposing an interconnected structure with three crucial blocks: noise modeling, source knowledge identification, and enhanced noise detection using noise source-knowledge-integration methods.
	CHAMMI: A benchmark for channel-adaptive models in microscopy imaging Zitong Chen, Chau Pham, Siqi Wang, Michael Doron, Nikita Moshkov, Juan C. Caicedo, Bryan A. Plummer, Advances in Neural Information Processing Systems (NeurIPS), 2023 arXiv We present a benchmark for investigating channel-adaptive models in microscopy imaging, which consists of 1) a dataset of varied-channel single-cell images, and 2) a biologically relevant evaluation framework. In addition, we adapted several existing techniques to create channel-adaptive models and compared their performance on this benchmark to fixed-channel, baseline models.
	Anchoring to Exemplars for Training Mixture-of-Expert Cell Embeddings Siqi Wang, Manyuan Lu, Nikita Moshkov, Juan C. Caicedo, Bryan A. Plummer, arXiv, 2021 arXiv Treatment ExemplArs with Mixture-of-experts (TEAMs), an embedding learning approach that learns a set of experts that are specialized in capturing technical variations in our training set and then aggregates specialist's predictions at test time.
	Stylistic Compatibility Learning with Deep Neural Networks for Indoor Scene Siqi Wang, Daniel Ritchie (advisor), Master Project Report, 2020 A deep neural network with conditioning method to learn the scene style.
	An Efficient Adaptive Algorithm for Removal of Impulse Noises Siqi Wang, Tongyu Yue, Bo Lang, International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI), 2017 Adaptive Min-Max Average Filters (AMMAF) for the removal of impulse noises.
	Efficient Segmentation for Region-based Image Retrieval Using Edge Integrated Minimum Spanning Tree Yang Liu, Lei Huang, Siqi Wang, Xianglong Liu, Bo Lang, International Conference on Pattern Recognition (ICPR), 2016 A RBIR-oriented image segmentation algorithm named Edge Integrated Minimum Spanning Tree (EI-MST).
Template from source code.