Learning and Inferring Transportation Routines. L. Liao, D. Fox, and H. Kautz. Proceedings of the Nineteenth National Conference on Artificial Intelligence (AAAI-2004), San Jose, CA, 2004.
The paper shows how the variable domains and parameters of a hierarchically-structured Dynamic Bayesian Network can be learned from unlabeled data. The hierarchical structure captures activity at the level of goals (destinations), plans (trip segments), and actions (movements on the street grid). The paper also shows that novel detection (sometimes called abnormality detection) can be implemented by conditional likelihood of the data against the trained and untrained models.
Extracting Places and Activities from GPS Traces Using Hierarchical Conditional Random Fields. L. Liao, D. Fox, and H. Kautz. International Journal of Robotics Research, 26:105-117, 2007.
This paper provides a good introduction to inference and parameter learning in conditional random fields, using the problem of interpreting raw GPS input in terms of transportation activities such as "walk" and "ride a bus", and non-transportation activities that are are performed at "significant places", such as "shop" or "visit friend". Processing proceeds in stages. First, a CRF is used to correct for GPS error, resulting in a smooth path along streets. In the second stage, the algorithm iteratively infers activitiy labels and infers the existence of significant places. This iterative method illustrates one solution to the general problem of building a CRF when the number of nodes is not known in advance. Lin Liao's C++ Conditional Random Field Toolkit can be downloaded here.
Understanding Belief Propagation and Its Generalizations. J. Yedidia, W. Freeman, and Y. Weiss. In Exploring Artificial Intelligence in the New Millennium, Chap. 8, pp. 239-236, Morgan Kaufmann, 2003.
"Inference" problems arise in statistical physics, computer vision, error-correcting coding theory, and AI. We explain the principles behind the belief propagation (BP) algorithm, which is an efficient way to solve inference problems based on passing local messages. We develop a unified approach, with examples, notation, and graphical models borrowed from the relevant disciplines.We explain the close connection between the BP algorithm and a detailed example.
Markov Logic, P. Domingos, S. Kok, D. Lowd, H. Poon, M. Richardson and P. Singla. In L. De Raedt, P. Frasconi, K. Kersting and S. Muggleton (eds.), Probabilistic Inductive Logic Programming. New York: Springer, 2008.
Inference, learning, and applications for Markov Logic networks. Includes slice sampling and discriminitive learning. A good place to start. Download Alchemy from here.
Hybrid Markov Logic Networks, J. Wang and P. Domingos. Proceedings of the Twenty-Third National Conference on Artificial Intelligence, Chicago, IL, 2008.
Extends Markov Logic to include real-valued variables. Download the data sets used in the paper from here.
Lifted First-Order Belief Propagation, P. Singla and P. Domingos. Twenty-Third National Conference on Artificial Intelligence, Chicago, IL, 2008.
Methods for avoiding instantiating an entire Markov Logic Network by reasoning about sets of similar entities.
Discovery of Social Relationships in Consumer Photo Collections using Markov Logic, P. Singla, H. Kautz, J. Luo and A. Gallagher. CVPR Workshop on Semantic Learning and Applications in Multimedia, Anchorage, Alaska, 2008.
Example of using Markov Logic to model relationships between individuals in family photo albums.
Learning Hidden Curved Exponential Random Graph Models to Infer Face-to-Face Interaction Networks from Situated Speech Data. D. Wyatt, T. Choudhury, and J. Bilmes. Proceedings of the Twenty-third National Conference on Artificial Intelligence (AAAI 2008), Chicago, IL, 2008.
Recent work on inferring a social network from data (in this case, information about conversations). Good citations to the literature on inferring and modeling social networks, and valuable discussion of the problem of inferring the network structure when data is noisy.