Kezhen Chen's Homepage
Research Scientist
(Google) X, the moonshot factory
Computer Science Ph.D, Northwestern University 2016 - 2022
-
Building a hybrid vision-language system that can perform complicated reasoning ability to multiple tasks
-
Developed a human-like object recognition model based on qualitative representations and analogical reasoning
-
Designed a pipeline for human-like action recognition on the skeleton data
Research Intern, Microsoft Research, Redmond 2020 Summer
-
Building a understandable deep learning model using Tensor Product Representation (TPR) and symbolic reasoning.
-
Developing new vision and language task "Image commenting" for generating emotional comments on images.
Research Engineer (Vendor), Microsoft Research, Redmond 2019.10 - 2020.5
-
Creating new vision and language dataset and task "Image commenting".
Research Intern, Microsoft Research, Redmond 2019 Summer
-
Building a neural-symbolic deep learning model for solving math problems via generating a sequence of equations.
Research Assistant, Vista AI Lab, University of Rochester 2013 - 2016
-
Analyzed the fashion trend via Computer Vision algorithm on images from social media, advised by Prof. Jiebo Luo
-
Worked in Twitter Health project to predict flu trend from tweets, advised by Prof. Henry Kautz
Research Assistant, HCI Lab, University of Rochester 2015 - 2016
-
Built an HCI system that analyzed people's body language during speech and gave improvement feedback
-
Conducted experiments that evaluated the speech feedback system
EDUCATION
RESEARCH INTERESTS
Machine Learning
I am interested in developing and using machine learning and deep learning models on various tasks.
2016 - Present
Northwestern University
Ph.D. in Computer Science.
Qualitative Representation and Reasoning
I am interested in modeling human-like symbolic qualitative representations&reasoning, analogical reasoning, case-based reasoning.
Vision-Language
I am interested in working on vision-language tasks with both deep learning methods and symbolic reasoning.
Neural-Symbolic
I am interested in taking advantage of tradition symbolic reasoning and deep learning. I aim to design deep learning models combining traditional symbolic reasoning as inductive bias.
2012 - 2016
University of Rochester
Bachelor's Degree in Computer Science