Hello

My name is Bryan Lu. I am a master's student in the machine learning department at Carnegie Mellon University, working with Katia Sycara. Previously, I graduated from Vanderbilt University with honors in computer science and mathematics. I interned at AWS Titan Labs and LLNL Machine Intelligence Group.
I have a diverse background in machine learning. Starting my research journey from deep learning applications in medical imaging & data visualization, I was deeply intrigued by the capability of these neural machinaries. Since then, I have been exploring tools / theories to systematically understand them. Ultimately, I want to build safe and responsible technologies.
As a first-generation student, I believe everyone should have access to quality education and dare to grow, no matther their background. I co-founded Gaokaopedia. I am currently part of AI Mentoring for Underrepresented Student Groups at CMU.

Selected Publications

Characterizing Out-of-Distribution Error via Optimal Transport
Yuzhe Lu*, Yilong Qin*, Runtian Zhai, Andrew Shen, Ketong Chen, Zhenlin Wang, Soheil Kolouri, Simon Stepputtis, Joseph Campbell, Katia Sycara
Estimating a classifier's performance on OOD data with Optimal Transport with theoretical gaurantees.
Conference on Neural Information Processing Systems (NeurIPS). 2023.
Project PDF

Set Locality Sensitive Hashing via Sliced Wasserstein Embeddings
Yuzhe Lu*, Xinran Liu*, Andrea Soltoggio, Soheil Kolouri
Fast set data retrieval through sliced Wassersten embedding and non-parametric learning.
Winter Conference on Applications of Computer Vision (WACV) (submitted). 2022.
Project PDF

Wasserstein Task Embedding for Measuring Task Similarities
Xinran Liu, Yikun Bai, Yuzhe Lu, Andrea Soltoggio, Soheil Kolouri
Highly efficient and model agnostic comparison of machine learning tasks.
Neural Networks (submitted). 2022.
PDF

An Interactive Interpretability System for Breast Cancer Screening with Deep Learning
Yuzhe Lu, Adam Perer
A visual analytics interface empowering radiologists to identify interpretable units in deep convolutional neural networks at scale to generate customized explanations.
IEEE Visualization Conference (IEEE VIS). 2022.
Project PDF

Scale-reinforced Implicit Neural Representations
Saroj Sahoo, Yuzhe Lu, Matthew Berger
Our approach accepts as input an arbitrary coordinate-based neural network modeling an underlying field and reinforces the network with scale, reproducing a given scale-parameterized family of transformations.
IEEE Transactions on Visualization and Computer Graphics (IEEE TVCG) (submitted). 2022.

Neural Flow Map Reconstruction
Saroj Sahoo, Yuzhe Lu, Matthew Berger
A reconstruction technique for the reduction of unsteady flow data based on neural representations of time-varying vector fields.
Computer Graphics Forum (EuroVis Proceedings). 2022.
PDF

Compressive Neural Representations of Volumetric Scalar Fields
Yuzhe Lu, Kairong Jiang, Joshua A. Levine, Matthew Berger
A state-of-the-art volumetric scalar fields compression method using implicit neural representations.
Computer Graphics Forum (EuroVis Proceedings). 2021.
Project PDF

Holistic Fine-grained Global Glomerulosclerosis Characterization: From Detection to Unbalanced Classification
Yuzhe Lu, Haichun Yang, Zuhayr Asad, Zheyu Zhu, Tianyuan Yao, ..., Yuankai Huo
A holistic pipeline to perform fine-grained glomeruli sclerosis quantification directly from high-resolution whole slide images.
Journal of Medical Imaging (JMI). 2021.
PDF

SimTriplet: Simple Triplet Representation Learning with a Single GPU
Quan Liu, Peter C. Louis, Yuzhe Lu, Aadarsh Jha, Mengyang Zhao, ..., Yuankai Huo
A contrastive learning framework taking advantage of the multi-view nature of medical images beyond self-augmentation.
International Conference on Med Image Computing and Computer-Assisted Intervention (MICCAI). 2021.
Project PDF

EasierPath: An Open-source Tool for Human-in-the-loop Deep Learning of Renal Pathology
Zheyu Zhu, Yuzhe Lu, Ruining Deng, Haichun Yang, Agnes B. Fogo, Yuankai Huo
An open-source, human-in-the-loop tool to integrate physicians and deep learning algorithms for large-scale pathological image quantification.
International Conference on Med Image Computing and Computer-Assisted Intervention (MICCAI). 2020.
Project PDF

CircleNet: Anchor-Free Glomerulus Detection with Circle Representation
Haichun Yang, Ruining Deng, Yuzhe Lu, Zheyu Zhu, Ye Chen, ..., Yuankai Huo
An anchor-free detection method using circle representation to detect the ball-shaped glomerulus.
International Conference on Med Image Computing and Computer-Assisted Intervention (MICCAI). 2020.
Project PDF