This Article 
 Bibliographic References 
 Add to: 
Learning Graphical Model Parameters with Approximate Marginal Inference
Oct. 2013 (vol. 35 no. 10)
pp. 2454-2467
J. Domke, NICTA, Australia Nat. Univ., Canberra, ACT, Australia
Likelihood-based learning of graphical models faces challenges of computational complexity and robustness to model misspecification. This paper studies methods that fit parameters directly to maximize a measure of the accuracy of predicted marginals, taking into account both model and inference approximations at training time. Experiments on imaging problems suggest marginalization-based learning performs better than likelihood-based approximations on difficult problems where the model being fit is approximate in nature.
Index Terms:
solid modelling,approximation theory,computational complexity,inference mechanisms,learning (artificial intelligence),likelihood-based approximations,graphical model parameter learning,approximate marginal inference,likelihood-based learning,computational complexity,model misspecification,inference approximations,marginalization-based learning,Vectors,Entropy,Approximation algorithms,Optimization,Function approximation,Markov processes,segmentation,Graphical models,conditional random fields,machine learning,inference
J. Domke, "Learning Graphical Model Parameters with Approximate Marginal Inference," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 35, no. 10, pp. 2454-2467, Oct. 2013, doi:10.1109/TPAMI.2013.31
Usage of this product signifies your acceptance of the Terms of Use.