Probabilistic Reasoning using Graphical Models

In probabilistic graphical models, we have significant experience in developing theories for graphical model learning and efficient inference for different graphical models including both directed and undirected models. In graphical model learning, we have developed algorithms for learning the model parameters and structure, and learning the model under incomplete training data with qualitative constraints. Using directed models, we also did some work on feature learning. For inference, we focus on developing methods that can perform efficient and timely inference by factoring out the common computations and by identifying the most informative evidences to use. We also developed a unified probabilistic framework based on combining the directed and undirected graphs to produce a more expressive framework that can capture heterogeneous knowledge at different levels of abstraction.

1. Theoretical Developments

Over the past few years, we have developed several techniques related to learning and inference in directed, undirected and hybrid models.

(1) Bayesian Network Learning

(2) Bayesian Network Structure Learning with Bounded Treewidth

(3) Regression Bayesian Networks

(4) Local Structure Learning and Causal Discovery

(5) Efficient Inference

(6) Active Information Fusion for Decision Making

(7) Effects-based Planning

(8) Learning and Inference in Hybrid Graphical Model

(9) Undirected Models


Some of the related papers can be downloaded here.

2. Applications

We have applied graphical models to different applications including human activity modeling and recognition, facial activity modeling and understanding, image and video segmentation, military planning, and information fusion for decision making and situation awareness. PGM Application projects include:

(1) Human Activity Modeling and Recognition (Sponsored by DARPA, Army)

(2) Facial Activity Modeling and Understanding (Sponsored by DARPA and ONR)

(3) Dynamic and Active Information Fusion for Decision Making Under Uncertainty (Sponsored by AFOSR and ARO)

(4) Efficient Effects-based Military Planning (sponsored by ARO)

(5) Image Segmentation

3. Softwares

(1) Our Software

The software for Bayesian network structure learning can be downloaded here.

(2) Other Software Packages

Some useful software packages for graphical models can be found here.

A good R package for BN structure learning can be found here.

A library of causal discovery algorithms based on Bayesian Network learning theory: CausalExplorer.

Another package for graphical models in R: CRAN Task.

Most recent softwares for BN structure learning: GOBNILP, URLearning, best-w-tree, TWILP, SparsityBoost, OpenGM.

4. Additional Resources

Some useful online resources for PGM. [link]

5. Publication

Journal

  • Jixu Chen, Siqi Nie, and Qiang Ji. Data-free Prior Model for Upper Body Pose Estimation and Tracking, IEEE Transactions on Image Processing (TIP), vol. 22, issue 12, pp. 4627-4639, 2013.

  • Yongmian Zhang, Yifan Zhang, Eran Swears, Natalia Larios, Ziheng Wang, and Qiang Ji. Modeling Temporal Interactions with Interval Temporal Bayesian Networks for Complex Activity Recognition, IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), Vol. 35, No. 10, 2013.

  • Lei Zhang, Zhi Zeng and Qiang Ji, Probabilistic Image Modeling with an Extended Chain Graph for Human Activity Recognition and Image Segmentation, IEEE Transactions on Image Processing (TIP), pages 2401-2413, Vol. 20, No. 9, September 2011.

  • Lei Zhang and Qiang Ji, A Bayesian Network Model for Automatic and Interactive Image Segmentation, IEEE Transactions on Image Processing (TIP), pages 2582-2593, Vol.20, No.9, September 2011.

  • Cassio de Campos and Qiang Ji, Efficient Structure Learning of Bayesian Networks using Constraints, Journal of Machine Learning Research (JMLR) 12, p663-689, 2011. [PDF] [Software]

  • Wenhui Liao and Qiang Ji, Learning Bayesian Network Parameters Under Incomplete Data with Qualitative Domain Knowledge, Pattern Recognition, Volume 42, Issue 11, p3046-3056, 2009. [PDF]

  • Conference

  • Yue Wu, Ziheng Wang and Qiang Ji. A Hierarchical Probabilistic Model for Facial Feature Detection, Computer Vision and Pattern Recognition (CVPR), 2014.

  • Yue Wu and Qiang Ji, Learning the Deep Features for Eye Detection in Uncontrolled Conditions, 22nd International Conference on Pattern Recognition (ICPR), 2014.

  • Siqi Nie and Qiang Ji. Capturing Global and Local Dynamics for Human Action Recognition, 22nd International Conference on Pattern Recognition (ICPR), 2014. (ORAL)

  • Siqi Nie and Qiang Ji. Feature Learning using Bayesian Linear Regression Model, 22nd International Conference on Pattern Recognition (ICPR), 2014.

  • Ziheng Wang, Shangfei Wang and Qiang Ji. Capturing Global Semantic Relationships for Facial Action Unit Recognition, International Conference on Computer Vision (ICCV), 2013.

  • Yue Wu, Zuoguan Wang and Qiang Ji, Facial Feature Tracking under Varying Facial Expressions and Face Poses based on Restricted Boltzmann Machines, Computer Vision and Pattern Recognition (CVPR), 2013.

  • Ziheng Wang, Shangfei Wang and Qiang Ji. Capturing Complex Spatio-Temporal Relations among Facial Muscles for Facial Expression Recognition, Computer Vision and Pattern Recognition (CVPR), 2013.

  • Zuoguan Wang, Siwei Lyu, Gerwin Schalk, and Qiang Ji. Deep Feature Learning using Target Priors with Applications in ECoG Signal Decoding for BCI. International Joint Conferences on Artificial Intelligence (IJCAI), 2013. (Oral)

  • Zuoguan Wang, Siwei Lyu, Gerwin Schalk, and Qiang Ji. Learning with Target Prior. Annual Conference on Neural Information Processing Systems (NIPS), 2012.

  • Zuoguan Wang, Gerwin Schalk, and Qiang Ji. Anatomically Constrained Decoding of Finger Flexion from Electrocorticographic Signals. Annual Conference on Neural Information Processing Systems (NIPS), 2011. Spotlight presentation (top 5%).

  • Cassio de Campos and Qiang Ji, Properties of Bayesian Dirichlet scores to learn Bayesian network structures, 24th AAAI Conference on Artificial Intelligence (AAAI), 2010. [PDF]

  • Cassio de Campos, Zhi Zeng, Qiang Ji, An Improved Structural EM to Learn Dynamic Bayesian Nets, International Conference on Pattern Recognition (ICPR), 2010. [PDF]

  • Cassio de Campos, Zhi Zeng, and Qiang Ji, Structure Learning of Bayesian Networks using Constraints, International Conference on Machine Learning (ICML), 2009. [PDF]

  • Cassio de Campos and Qiang Ji, Improving Bayesian Network Parameter Learning using Constraints, International Conference in Pattern Recognition (ICPR), 2008. [PDF]

  • Wenhui Liao and Qiang Ji, Exploiting Qualitative Domain Knowledge for Learning Bayesian Network Parameters with Incomplete Data, International Conference in Pattern Recognition (ICPR), 2008. [PDF]

  • Yan Tong and Qiang Ji, Learning Bayesian Networks with Qualitative Constraints, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2008. [PDF]