Research

Beauty of Math

Hoeffding's Inequality and MartingalesDec 2022

In this article we present the Hoeffding's Inequality and its proof. To do so, we first go through the Hoeffding's Lemma and Markov's Inequality. Also, we introduce the concept of Martingales and related Azuma-Hoeffding inequality. [PDF]

Research Experience

Bandit ProblemSep 2022 - present

Supervised by Prof. Lifeng Lai, UC Davis.

Graduate Student Researcher in Lai's group.

  • Doing research on mathematical theory of Reinforcement Learning: bandit problem.
  • Study the regret bound of bandit algorithms, e.g. Exp3, in non-stationary environment, and design strategies to attack reward signal.

Edge AI System for Smart HomeJun 2021 - Sep 2021

Supervised by Prof. Xiaofan (Fred) Jiang, Columbia University.

Summer Research Assistant in Columbia Intelligent and Connected Systems Lab.

  • A system able to auto-discover the sensors connected to it and allocate sensor and actuator resources according to the smart home applications users specified, deployed in multiple lab rooms.
  • Survey on current Edge AI systems and Digital Twin (DT) representation, and design our own system architecture from MCU and sensor hardware to user application. Build the testbed from scratch.
  • Testbed: Arduino Uno reads custom sensor data (vibration, sound, etc.) and transmit them to ESP8266 via UART. ESP8266 packs data with DT format and MQTT discovery format, and sends them to lab computer server (HTTP with Flask) and Home Assistant Raspberry Pi server (MQTT). Integrate Google Nest camera into HA and get the video stream. Data stored in PostgreSQL and MarialDB.
Images Images

Fever Screening SystemJan 2020 - May 2021

Supervised by Prof. Xiaofan (Fred) Jiang, Columbia University.

Research Intern in Columbia Intelligent and Connected Systems Lab.

  • A low-cost system based on RGB-thermal camera for continuous fever screening of multiple people without human interaction, deployed in a restaurant and hospital clinic in New York City.
  • Investigate solutions. Match multiple RGB and thermal heads, train and deploy YOLOV3 (head detection) and FSA- Net (head orientation regression), estimate distance using non-identical RGB and thermal camera, etc.
  • Contribute to paper writing. Two publications accepted by SenSys 2020 Poster [ACM DL] and CPHS 2021 [ACM DL].
  • News: Columbia Electrical Engineering. Project website: ICSL Fever Screening.
Images

Khameleon Scheduler in Reinforcement LearningJul 2020 - Sep 2020

Supervised by Prof. Eugene Wu, Columbia University.

Research Intern in WuLab Columbia University.

  • The server-side scheduler involves a complex optimization based on available resources, predicted user interactions, and response quality levels to maximize user-perceived interactivity in real-time.
  • Create the simulated RL environment, which can pre-compute the transition relationships or compute it dynamically in run-time.
  • Write Q-Learning and SARSA based prefetching scheduler to trade off latency for response quality with the progressive encoded response in cloud-based interactive applications.
  • Add predicted user preferred action choice to algorithms above.
  • Project GitHub: Khameleon-Scheduler-RL.
Images Images

Optimization of Integrated OpticsFeb 2019 – Jun 2019

Supervised by Prof. Xiaoqi Zhou, Sun Yat-sen University.

Research Intern in Optical Quantum Information Lab.

  • Undergraduate thesis.
  • Design numerical methods to solve optimal parameters of grating coupler based on regression analysis and constrained optimization problem.
  • Conduct simulation experiment with one-dimensional grating coupler.
  • Outcome: One thesis paper written.

Cyber-Physical Energy SystemsOct 2017 - Feb 2019

Supervised by Prof. Jiang Wu, Xi'an Jiaotong University.

Member of XJTU Information-technology Talent Program.

Research Intern in Ministry of Education Key Lab for Intelligent Networks and Network Security.

  • Propose centralized K-means algorithm, using weighted combination of features from PCA and prior knowledge. Expand it to distributed cases, using parameter consensus and feature transferring method.
  • Write centralized & distributed K-means clustering algorithms for massive data on energy demand side. Analyze the results obtained from the power system in a city. Contribute to paper writing.
  • Outcome: Two papers written and one accepted by Chinese Control Conference [DOI].
Images

Publication

University of California, Davis

  1. Chenye Yang, Guanlin Liu, Lifeng Lai, "Reward Attack on Stochastic Bandits with Non-stationary Rewards" (Under review for Asilomar 2023).

Columbia University

  1. Kaiyuan Hou, Yanchen Liu, Peter Wei, Chenye Yang, Hengjiu Kang, Stephen Xia, Teresa Spada, Andrew Rundle, Xiaofan Jiang, "A Low-Cost In-situ System for Continuous Multi-Person Fever Screening" (IPSN 2022) [IEEE DOI].
  2. Stephen Xia, Rishikanth Chandrasekaran, Yanchen Liu, Chenye Yang, Tajana Simunic Rosing, Xiaofan Jiang, "A Drone-based System for Intelligent and Autonomous Homes" (SenSys 2021 Best Demo Award) [ACM DL].
  3. Peter Wei, Yanchen Liu, Hengjiu Kang, Chenye Yang, Xiaofan Jiang, "A Low-Cost and Scalable Personalized Thermal Comfort Estimation System in Indoor Environments" (CPHS 2021) [ACM DL].
  4. Peter Wei, Chenye Yang, Xiaofan Jiang, "Low-Cost Multi-Person Continuous Skin Temperature Sensing System for Fever Detection: Poster Abstract" (SenSys 2020 Poster) [ACM DL].

Sun Yat-sen University

  1. Chenye Yang, "Optimal Design of Integrated Photonic Devices Based on Regression Analysis and Constrained Optimization Problem Solving" (Xi'an Jiaotong University undergraduate thesis).

Xi'an Jiaotong University

  1. Pengyuan Liu, Chenye Yang, Jiang Wu, "Hybrid Features Based K-means Clustering Algorithm for Use in Electricity Customer Load Pattern Analysis" (Chinese Control Conference 2018) [IEEE DOI].