Han Zhao | 赵晗


PhD student
Machine Learning Department
School of Computer Science
Carnegie Mellon University
Email: han (DOT) zhao [AT] cs (DOT) cmu (DOT) edu
Office: GHC 8021
[Curriculum Vitae] [Google Scholar] [Research Statement]

Bio

I am a final-year PhD student at the Machine Learning Department, Carnegie Mellon University. I am fortunate to work with my advisor, Prof. Geoff Gordon. Before coming to CMU, I obtained my BEng degree from the Computer Science Department at Tsinghua University and MMath from the University of Waterloo. I have a broad interest in both the theoretical and applied side of machine learning. In particular, I work on invariant representation learning, probabilistic reasoning with Sum-Product Networks, transfer and multitask learning, and computational social choice.

I also co-organize the AI seminar @ CMU, feel free to drop me an email if you are interested to give a talk!

Publications [ show selected / show by date / show by topic ]

Invariant Representation Learning

My work proves a fundamental limitation of learning invariant representations. With this result, we also identify and explain the inherent trade-offs in unsupervised domain adaptation and algorithmic fairness. Selected papers include:
Conditional Learning of Fair Representations
H. Zhao, A. Coston, T. Adel and G. Gordon
In Proceedings of the 8th International Conference on Learning Representations (ICLR 2020, Spotlight)
NeurIPS 2019 Workshop on Machine Learning with Guarantees (NeurIPS 2019)
[abs] [pdf] [video] [slides] [code]
Inherent Tradeoffs in Learning Fair Representations
H. Zhao and G. Gordon
In Proceedings of the 33rd Advances in Neural Information Processing Systems (NeurIPS 2019)
[abs] [pdf] [poster] [slides] [blog]
On Learning Invariant Representations for Domain Adaptation
H. Zhao, R. Tachet, K. Zhang and G. Gordon
In Proceedings of the 36th International Conference on Machine Learning (ICML 2019, Long Oral)
[abs] [pdf] [supplement] [poster] [slides] [blog]
Adversarial Multiple Source Domain Adaptation
H. Zhao*, S. Zhang*, G. Wu, J. Costeira, J. Moura and G. Gordon
In Proceedings of the 32nd Advances in Neural Information Processing Systems (NeurIPS 2018)
[abs] [pdf] [supplement] [poster] [code]

Tractable Probabilistic Reasoning

My work establishes the equivalence between Sum-Product networks, Bayesian networks with algebraic decision diagrams, and mixture models with exponentially many components. With these theoretical results, we propose efficient learning algorithms for Sum-Product networks in both offline and online, distributed and Bayesian settings. Selected papers include:
A Unified Approach for Learning the Parameters of Sum-Product Networks
H. Zhao, P. Poupart and G. Gordon
In Proceedings of the 30th Advances in Neural Information Processing Systems (NIPS 2016)
[abs] [pdf] [supplement] [poster] [code]
Collapsed Variational Inference for Sum-Product Networks
H. Zhao, T. Adel, G. Gordon and B. Amos
In Proceedings of the 33rd International Conference on Machine Learning (ICML 2016)
[abs] [pdf] [poster] [slides] [code]
On the Relationship between Sum-Product Networks and Bayesian Networks
H. Zhao, M. Melibari and P. Poupart
In Proceedings of the 32nd International Conference on Machine Learning (ICML 2015)
[abs] [pdf] [supplement] [Full arXiv version] [slides] [poster]

Workshop Papers and Pre-prints

Adversarial Privacy Preservation under Attribute Inference Attack
H. Zhao*, J. Chi*, Y. Tian and G. Gordon
NeurIPS 2019 Workshop on Machine Learning with Guarantees (NeurIPS 2019)
[abs] [pdf]
Approximate Empirical Bayes for Deep Neural Networks
H. Zhao*, Y. H. Tsai*, R. Salakhutdinov and G. Gordon
In Uncertainty in Deep Learning workshop at UAI (UAI UDL 2018)
[abs] [pdf] [poster]
Multiple Source Domain Adaptation with Adversarial Learning
H. Zhao*, S. Zhang*, G. Wu, J. Costeira, J. Moura and G. Gordon
In 6th International Conference on Learning Representations (ICLR 2018 workshop track)
[abs] [pdf] [poster]
Discovering Order in Unordered Datasets: Generative Markov Networks
Y. H. Tsai, H. Zhao, R. Salakhutdinov and N. Jojic
In Time Series workshop at NIPS (NIPS TSW 2017)
[abs] [pdf] [slides] [poster]
A Sober Look at Spectral Learning
H. Zhao and P. Poupart
In Method of Moments and Spectral Learning workshop at ICML (ICML MM 2014)
[abs] [pdf] [slides] [poster] [code]

Misc

I enjoy sketching and calligraphy at my spare time. If I have a long vacation, I also enjoy traveling.