Research Scientist

Google Brain, Mountain View

Email: bodai AT google.com

Short Bio

I am a research scientist in Google Brain. I obtained my Ph.D. from Computational Science and Engineering at Georgia Tech.

My principal research interest lies on principled machine learning for structured data. I am particularly focusing on developing effective statistical models and efficient algorithms which learns from a massive volume of complex, structured, uncertain and high-dimensional data, e.g., distributions, structures, dynamics, and so on.

My research work includes three major themes:

  • Reinforcement learning: design effective algorithms by exploiting the intrinsic structures in the uncertain dynamics for automatic decision making.
  • Structured input and output: build effective models for capturing the structures information in input and output, e.g., binaries, sequences, programs, trees, and graphs.
  • Large-scale nonparametric machine learning: develop efficient algorithms for machine learning methods, especially nonparametric methods, to handle hundreds of millions of data.

More information can be found in my Google AI page.


  • 2018/09: Four papers have been accepted to NIPS2018.
  • 2018/08: I will join Google Brain as a research scientist.
  • 2018/05: Three papers have been accepted to ICML2018.
  • 2018/05: One paper has been accepted to UAI2018.
  • 2018/03: One paper has been accepted to IJCAI-ECAI2018.
  • 2018/02: One papers has been accepted to CVPR2018.
  • 2018/01: Two papers have been accepted to ICLR2018.
  • 2017/12: One paper has been accepted to AISTATS2018.
  • 2017/12: Our paper, "Syntax-Directed Variational Autoencoder for Molecule Generation", won the Best Paper Award in NIPS2017 Machine Learning for Molecules and Materials workshop .
  • 2017/09: One paper has been accepted to NIPS2017.
  • 2017/05: Two papers have been accepted to ICML2017.

Selected Publications

  • Bo Dai*, Hanjun Dai*, Niao He, Weiyang Liu, Zhen Liu, Jianshu Chen, Lin Xiao, Le Song. "Coupled Variational Bayes via Optimization Embedding". Neural Information Processing Systems (NIPS'2018).
  • Bo Dai, Albert Shaw, Lihong Li, Lin Xiao, Niao He, Zhen Liu, Jianshu Chen, Le Song. "SBEED: Convergent Reinforcement Learning with Nonlinear Function Approximation". The 35th International Conference on Machine Learning (long talk, ICML'2018).
  • Bo Dai*, Albert Shaw*, Niao He, Lihong Li, Le Song. "Boosting the Actor with Dual Critic". The 6th International Conference on Learning Representations (ICLR'2018).
  • Bo Dai, Niao He, Yunpeng Pan, Byron Boots, and Le Song. "Learning from Conditional Distributions via Dual Embeddings". The 20th International Conference on Artificial Intelligence and Statistics (AISTATS'2017).
  • Hanjun Dai, Bo Dai and Le Song. "Discriminative Embeddings of Latent Variable Models for Structured Data". The 33th International Conference on Machine Learning (ICML'2016). [CODE]
  • Bo Dai, Niao He, Hanjun Dai and Le Song. "Provable Bayesian Inference via Particle Mirror Descent", The 19th International Conference on Artificial Intelligence and Statistics (Best Student Paper Award, full oral presentation, AISTATS'2016).
  • Bo Dai, Bo Xie, Niao He, Yingyu Liang, Anant Raj, Maria-Florina Balcan and Le Song. "Scalable Kernel Methods via Doubly Stochastic Gradients", Neural Information Processing Systems (NIPS'2014). [CODE]