My primary research interest is the foundation of statistical inference in machine learning (ML). I am particularly interested in computationally efficient and scalable generic inference algorithms beyond Markov chain Monte Carlo and variational inference.
I am also interested in many ML models that requires start-of-the-art inference algorithms. Recently, I am working on efficient training algorithm on Boltzmann machines, where the network architecture is beyond of the scope of deep learning. The applications of this kind of Boltzmann machines on unsupervised continual learning are particularly interesting to me.
When I started my PhD in 2010, I could not expect that our everyday life could be shaped by ML technologies like now. I believe that machine learning research for data science is a very exciting new research direction. The scope of my research interest on data science is rather broad, which covers from best practice of ML technologies in production to the challenges in data science team project management. My current research in this area is focused on smart data preprocessing to help data scientists and ML engineers to handle most challenging data issues.
The Theory and Algorithm of Ergodic Inference. Yichuan Zhang. arxiv.org 2018.
Ergodic Measure Preserving Flows. Yichuan Zhang, Jose Miguel Hernandez-Lobato, Zoubin Ghahramani. arxiv.org 2018.
Semi-Separable Hamiltonian Monte Carlo for Inference in Bayesian Hierarchical Models. Yichuan Zhang, Charles Sutton. In Advances in Neural Information Processing Systems (NIPS). 2014.
Continuous Relaxations for Discrete Hamiltonian Monte Carlo. Yichuan Zhang, Charles Sutton. In Advances in Neural Information Processing Systems (NIPS). 2012.
Quasi-Newton Markov chain Monte Carlo. Yichuan Zhang, Charles Sutton. In Advances in Neural Information Processing Systems (NIPS). 2011.
Inference on Boltzmann Machines-Probabilistic Relaxation, Hamiltonian Monte Carlo and Geometry. Yichuan Zhang. at CamAIML. 2018.
post-doc research associate
PhD Machine Learning
PhD Machine Learning