Associate Professor at the Khoury College of Computer Sciences, Northeastern University
Machine Learning Researcher specializing in Kernel Methods, Neural Networks, and their applications in interdisciplinary domains
Get in TouchMy research focuses on advancing machine learning theory and applications, with a specific emphasis on kernel methods and deep learning. I'm passionate about discovering a unifying mathematical theory for neural networks and exploring interdisciplinary applications of machine learning.
Pro. Wu is one of the original researchers on interpretable kernel methods. With a focus on the Hilbert Schmidt Independence Criterion (HSIC), Pro. Wu discovered the optimization technique of ISM (Iterative Spectral Method).
In 2022, Prof. Wu made the discovery that neural networks can have a closed-form solution by changing the activation function from ReLU to cosines. Prof. Wu is currently looking a a general closed form solution.
Applying machine learning to identify personalized risk factors and predict gestational age through a grant from the Opportunity and Infrastructure Fund.
Developing machine learning approaches for analyzing environmental data, including single-cell Raman spectroscopy for bacterial taxonomy identification.
Researching instance-wise feature grouping methodologies to enhance interpretability and performance of complex machine learning models.
Prof. Wu is currently collaborating with the Biology team to better understand the Lyme bacteria structure.
As a machine learning scholar, I've published papers in top conferences and journals, with a focus on kernel methods, deep learning theory, and interdisciplinary applications.
Abstract: There is currently a debate within the neuroscience community over the likelihood of the brain performing backpropagation (BP). To better mimic the brain, training a network one layer at a time with only a "single forward pass" has been proposed as an alternative to bypass BP; we refer to these networks as "layer-wise" networks. We continue the work on layer-wise networks by answering two outstanding questions. First, do they have a closed-form solution? Second, how do we know when to stop adding more layers? This work proves that the Kernel Mean Embedding is the closed-form weight that achieves the network global optimum while driving these networks to converge towards a highly desirable kernel for classification; we call it the Neural Indicator Kernel.
Abstract: In many learning problems, the domain scientist is often interested in discovering the groups of features that are redundant and are important for classification. Moreover, the features that belong to each group, and the important feature groups may vary per sample. But what do we mean by feature redundancy? In this paper, we formally define two types of redundancies using information theory: Representation and Relevant redundancies. We leverage these redundancies to design a formulation for instance-wise feature group discovery and reveal a theoretical guideline to help discover the appropriate number of groups. We approximate mutual information via a variational lower bound and learn the feature group and selector indicators with Gumbel-Softmax in optimizing our formulation. Experiments on synthetic data validate our theoretical claims. Experiments on MNIST, Fashion MNIST, and gene expression datasets show that our method discovers feature groups with high classification accuracies.
Abstract: This paper presents ensemble learning methods to identify key factors contributing to preterm births, with potential applications in preventive healthcare. Our work leverages a rich dataset collected by a NIEHS P42 Center that is trying to identify the dominant factors responsible for the high rate of premature births in northern Puerto Rico. We investigate analytical models addressing major challenges in this domain, using undersampling techniques combined with ensemble learning to improve identification of relevant risk factors.
Abstract: This research focuses on interpretable kernel dimensionality reduction methods, exploring how to make complex dimensionality reduction techniques more understandable and explainable. The paper presents novel approaches to solving kernel-based dimensionality reduction while maintaining interpretability, which is crucial for applications in domains where understanding the transformation process is as important as the reduction itself.
Abstract: Single-cell Raman Spectroscopy (SCRS) emerges as a promising tool for single-cell phenotyping in environmental ecological studies, offering non-intrusive, high-resolution, and high-throughput capabilities. In this study, we obtained a large and comprehensive SCRS dataset that captured phenotypic variations with cell growth status for various bacterial species. We apply machine learning approaches to analyze this data for bacterial taxonomy identification, demonstrating high accuracy in classifying bacterial species based on their Raman spectroscopy signatures.
Abstract: Rapid progress in various advanced analytical methods such as single-cell technologies enable unprecedented and deeper understanding of microbial ecology beyond the resolution of conventional approaches. A major application challenge exists in the determination of sufficient sample size without sufficient prior knowledge of the community complexity. This case study focuses on using single-cell Raman spectroscopy data in Enhanced Biological Phosphorus Removal (EBPR) systems, developing machine learning approaches to determine optimal sampling strategies for complex environmental systems.
I create educational videos to help students better understand complex machine learning concepts. These supplementary materials reinforce classroom learning and provide additional support for challenging topics.
I joined the Khoury College faculty in 2023 out of a love for teaching. My approach emphasizes deep understanding of machine learning concepts, balancing mathematical foundations with practical applications.
A comprehensive course covering fundamental machine learning algorithms, their mathematical principles, and practical implementations in Python. Students learn to develop a deep understanding of how these algorithms work "under the hood."
This graduate-level course explores cutting-edge research in machine learning, with a focus on kernel methods, deep learning theory, and emerging applications in interdisciplinary domains.
A rigorous exploration of the mathematical concepts underpinning modern machine learning algorithms, designed to provide students with a solid theoretical foundation.
I'm always open to research collaborations, speaking engagements, or questions from prospective students.
Email: ch.wu@northeastern.edu
Office: Khoury College of Computer Sciences, Northeastern University, Boston
When I'm not researching or teaching, I enjoy spending my time with various activities.
Enjoying team sports and staying active on the court.
A recent passion that combines strategy and physical activity.
Finding balance and creativity through music.