Advertisement

A Discriminative Model Corresponding to Hierarchical HMMs

  • Takaaki Sugiura
  • Naoto Goto
  • Akira Hayashi
Part of the Lecture Notes in Computer Science book series (LNCS, volume 4881)

Abstract

Hidden Markov Models (HMMs) are very popular generative models for sequence data. Recent work has, however, shown that on many tasks, Conditional Random Fields (CRFs), a type of discriminative model, perform better than HMMs. We propose Hierarchical Hidden Conditional Random Fields (HHCRFs), a discriminative model corresponding to hierarchical HMMs (HHMMs). HHCRFs model the conditional probability of the states at the upper levels given observations. The states at the lower levels are hidden and marginalized in the model definition. We have developed two algorithms for the model: a parameter learning algorithm that needs only the states at the upper levels in the training data and the marginalized Viterbi algorithm, which computes the most likely state sequences at the upper levels by marginalizing the states at the lower levels. In an experiment that involves segmenting electroencephalographic (EEG) data for a Brain-Computer Interface, HHCRFs outperform HHMMs.

Keywords

Hide Markov Model State Sequence Conditional Random Field Mental Task Viterbi Algorithm 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Lafferty, J.D., McCallum, A., Pereira, F.C.N.: Conditional random fields: Probabilistic models for segmenting and labeling sequence data. In: Proc. 18th Int. Conf. Machine Learning (2001)Google Scholar
  2. 2.
    Fine, S., Singer, Y., Tishby, N.: The hierarchical hidden Markov model: Analysis and applications. Machine Learning 32(1) (1998)Google Scholar
  3. 3.
    Murphy, K., Paskin, M.: Linear time inference in hierarchical HMMs. Advances in Neural Information Processing Systems 14 (2001)Google Scholar
  4. 4.
    Gotou, N., Hayashi, A., Suematsu, N.: Learning with segment boundaries for hierarchical HMMs. In: Proc. 3rd Int. Conf. Advances in Pattern Recognition (2005)Google Scholar
  5. 5.
    Sutton, C., McCallum, A., Rohanimanesh, K.: Dynamic conditional random fields: Factorized probabilistic models for labeling and segmenting sequence data. J. Mach. Learn. Res. 8 (March 2007)Google Scholar
  6. 6.
    Liao, L., Fox, D., Kautz, H.: Hierarchical conditional random fields for GPS-based activity recognition. In: Proc. 12th Int. Symp. of Robotics Research (2005)Google Scholar
  7. 7.
    Gunawardana, A., Mahajan, M., Acero, A., Platt, J.C.: Hidden conditional random fields for phone classification. In: Proc. Int. Conf. Speech Communication and Technology (2005)Google Scholar
  8. 8.
    Huang, C., Darwiche, A.: Inference in belief networks: A procedural guide. Int. J. of Approximate Reasoning 15(3) (1996)Google Scholar
  9. 9.
    Scheffer, T., Decomain, C., Wrobel, S.: Active hidden Markov models for information extraction. In: Hoffmann, F., Adams, N., Fisher, D., Guimarães, G., Hand, D.J. (eds.) IDA 2001. LNCS, vol. 2189, Springer, Heidelberg (2001)CrossRefGoogle Scholar
  10. 10.
    Obermaier, B., Guger, C., Neuper, C., Pfurtscheller, G.: Hidden markov models for online classification of single trial eeg data. Pattern Recogn. Lett. 22(12), 1299–1309 (2001)zbMATHCrossRefGoogle Scholar
  11. 11.
    Cincotti, F., et al.: Comparison of different feature classifiers for brain computer interfaces. In: Proc. 1st IEEE EMBS Conference on Neural Engineering, pp. 645–647 (2003)Google Scholar
  12. 12.
    Chiappa, S., Bengio, S.: Hmm and iohmm modeling of eeg rhythms for asynchronous bci systems. In: Proc. European Symposium on Artificial Neural Networks, pp. 199–204 (2004)Google Scholar
  13. 13.
    Blankertz, B., et al.: The bci competition iii: validating alternative approaches to actual bci problems. IEEE Transactions on Neural Systems and Rehabilitation Engineering 14(2), 153–159 (2006)CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2007

Authors and Affiliations

  • Takaaki Sugiura
    • 1
  • Naoto Goto
    • 1
  • Akira Hayashi
    • 1
  1. 1.Graduate School of Information Sciences, Hiroshima City University, 3-4-1 Ozuka-higashi, Asaminami-ku, Hiroshima 731-3194Japan

Personalised recommendations