Yang Liu


Yang Liu
Postdoctoral fellow
SEAS, CRCS, Harvard University
Office: MD-110

About me

    Hello. I'm currently a postdoctoral fellow at Harvard University. I am very fortunate to be hosted by Professor Yiling Chen, and collaborating with Professor David Parkes. I also affiliate with the Center for Research on Computation and Society. My research is broadly focused on the interactions between society and artificial intelligence (AI), and in particular algorithmic decision-making. Examples of topics that I enjoy researching on: machine learning aided incentive design (for collecting high quality data from human), crowdsourcing, algorithmic fairness in AI/ML, security/privacy for human data.

Upcoming workshop GaMeDATA18

  • [2018.04] We are thrilled to announce that Dave Pennock from MSR NYC will join us at GaMeDATA to talk about wagering mechanism!
  • [2018.02] We (together with Yiling Chen, Boi Faltings, David Parkes and Goran Radanovic) are organizing the first workshop on "Game-Theoretic Mechanisms for Data and Information" (GaMeDATA18) at the Federated AI Meeting of AAMAS, ICML, and IJCAI (FAIM) at Stockholm this incoming summer. Submissions of preliminary work, papers currently under review or in preparation for submission to other major venues in the field, and papers that have been accepted in relevant venues or published in the past year, are encouraged. Papers that appeared on arXiv, but haven’t been accepted for publication, are encouraged. The deadline is May 10. Feel free to reach out to me if there is any question about the workshop or the submission!


  • [2018.05] I'm happy to say that our paper “From Patching Delays to Infection Symptoms: Using Risk Profiles for an Early Discovery of Vulnerabilities Exploited in the Wild” is finally accepted to USENIX Security'18, after a long journey! We proved (empirically) that the cyber security world is well connected (the idea was first documented in my dissertation back to 2015, but Chaowei took it to another level) and this knowledge is useful! Stay tuned.
  • [2018.01] Our work (with Ji Liu and Tamer Başar) "Gossip Gradient Descent" is accepted to AAMAS'18 (extended abstract). We propose a O(1)-communication gossip algorithm that solves a distributed gradient descnet problem with strong theoretical guarantee.
  • [2017.11] We built a reinforcement learning framework for eliciting high quality information (reinforcement peer prediction). Preliminary version of this work is going to appear at Machine Learning in the Presence of Strategic Behavior at NIPS'17.
  • [2017.11] Our work (with CJ Ho) "Incentivizing High Quality User Contributions: New Arm Generation in Bandit Learning" is going to appear at AAAI'18.
  • [2017.06] Our paper "Calibrated Fairness in Bandits" is going to appear at FATML'17 (oral presentation, 8/48)!
  • [2017.06] Maintaining fairness in a decision making process: my paper "Fair Optimal Stopping Policy for Sequential Matching" is going to appear at UAI'17.

Selected / Recent Publications

<Click for the Full List>

Machine learning aided incentive design [Overview]

Surrogate Scoring Rules and a Dominant Truth Serum for Information Elicitation.
Yang Liu and Yiling Chen, in submission, 2017.

Machine Learning aided Peer Prediction.[Video] [Slides]
Yang Liu and Yiling Chen, ACM EC 2017, Cambridge, United States.

A Bandit Framework for Strategic Regression.[Video] [Poster]
Yang Liu and Yiling Chen, NIPS 2016, Barcelona, Spain.

[Artical] Designing More Informative Reputation Systems.
Yiling Chen, Jason Hartline, Yang Liu, Bo Waggoner and Dan Weld.

AI/ML for social good (fairness and security) [Overview]

Calibrated Fairness in Bandits (Fair Experimentation).[Slides]
Yang Liu, Goran Radanovic, Christos Dimitrakakis, Debmalya Mandal and David Parkes
FATML 2017, Halifax, Canada; also to appear at CODE@MIT 2017, Cambridge, United States.

Fair Optimal Stopping Policy for Sequential Matching.[Poster]
Yang Liu, UAI 2017, Sydney, Australia.

Cloudy with a Chance of Breach: Forecasting Cyber Security Incidents. [Slides]
Yang Liu, Armin Sarabi, Jing Zhang, Parinaz Ardabili, Manish Karir, Michael Bailey and Mingyan Liu
USENIX Security 2015, Washington, D.C., United States.

  • More details: Project webpage. You will also find a sample dataset to play with.
  • This work is featured in a WSJ Article.
  • QuadMetrics,Inc (acquired by FICO) , a start-up co-founded by Mingyan, is using the technique documented in the paper.
  • Patent: Rating Network Maliciousness and Comparing Network Maliciousness Through Similarity Analyses, No. 62/026,349.

Learning the wisdom of crowd [Overview]

An Online Learning Approach to Improving the Quality of Crowd-Sourcing.[Slides]
Yang Liu and Mingyan Liu, ACM SIGMETRICS 2015, Portland, United States.

Crowd Learning: Improving Online Decision Making Using Crowdsourced Data.[Slides]
Yang Liu and Mingyan Liu, IJCAI 2017, Melbourne, Australia.

[Tutorial] Bandit in Crowdsourcing.[Slides]
Yang Liu, CMO-BIRS Workshop: Models and Algorithms for Crowds and Networks 2016, Oxaca, Mexico.