Min Zheng

PhD Candidate, Health and AI Lab
Computer Science Department , Stevens Institute of Technology
mzheng3 at stevens dot edu

I'm a PhD candidate in the department of computer science at Stevens Institute of Technology, advised by Prof. Samantha Kleinberg , working in the Health and AI Lab (HAIL) at Stevens. My main research interests are causal inference and decision making modeling related to human health. I also work on machine learning projects that are related to automatic eating detection on data collected from multimodality sensors. I also have experience in building deep learning framework to detect change points in time series data.


Causal Inference and Causal Explanation

In this project, we focus on using causal inference and causal explanation methods to help understand why things happen. The data we focus on is time series data, especially in mdedical domain. Key application can be explanining why clinical events happen in medical dataset (E.g. Why a seizure happens and why a patient suffered from hypo/hyperglycemia). We proposed a method to explain why a token event occured (see detial in paper 1), and a method to identify causal moderators in time series using temporal logic rules (see detail in paper 2).

Related papers:
Paper 1: A Method for Automating Token Causal Explanation and Discovery. M. Zheng and S. Kleinberg.
FLAIRS, 2017. [pdf-link]
Paper 2: Automated Identification of Causal Moderators in Time-Series Data. M. Zheng, J. Claassen, and S. Kleinberg. ACM SIGKDD Causal Discovery Workshop, 2018. [pdf-link]

Causality and Decision Making

In this project, we focus on improving decision making using personalized causal models. We aim to develop methods that make the output of machine learning, specifically causal inference, useful for decision-making. In particular, we focus on challenging decisions where causal models may have the most benefit: when costs and rewards are at different timescales, sequential decisions that are linked over time, and cases where actions have complex and potentially uncertain effects.

Related papers:
[Two papers Under review for CogSCI 2019 and CRPI 2019]

Change Point Detection

In this project, we work on detect change points in time series data using deep learning method. We showed how changepoint detection can be treated as a supervised learning problem, and proposed a new deep neural network architecture that can efficiently identify both abrupt and gradual changes at multiple scales.

Related papers:
[Under review for IJCAI-2019]

Automatic Eating Detection

In this project, we focus on using machine learning to perform eating detection using body worn sensors including two Android watches for wrist motion, one in-ear microphone for audio data, and one Google Glass for head motion. We developed multi-modality eating detection methods to detect when, what, and how much people are eating in both lab and free-living environment.

Related papers:
Paper 1: Recognizing Eating from Body-Worn Sensors: Combining Free-living and Laboratory Data M. Mirtchouk, D. Lustig, A. Smith, I. Ching, M. Zheng, and S. Kleinberg.
IMWUT 1 (3) (previously UbiComp), 2017 [pdf-link]
Paper 2: Multimodality Sensing for Eating Recognition, C. Merck, C. Maher, M. Mirtchouk, M. Zheng, Y. Huang, and S. Kleinberg.
Pervasive Health, 2016 [pdf-link]

Meal Detection for Diabetes

In this project, we proposed a method to detect meals (including meal time and meal size) using CGM data from Type 1 Diabetes (T1D) patients. The correctly detected meals will allow insulin to be delivered in time, which helps T1D patients maintain healthy blood glucose.

Papers related:
[Preparing for submission]
The template of this website is based on the website template created by Dr. Adrian Dalca. Please check out his website for template details.