hmmTensor

R package for Hidden Markov Model (HMM) by Matrix/Tensor Decomposition

Installation

git clone https://github.com/rikenbit/hmmTensor/
R CMD INSTALL hmmTensor

or type the code below in the R console window

library(devtools)
devtools::install_github("rikenbit/hmmTensor")

Functions

• Standard HMM
  • Forward: Forward algorithm with scaling
  • Backward: Backward algorithm
  • Viterbi: Viterbi decoding (log-space)
  • BaumWelch: Baum-Welch EM algorithm (multi-sequence support)
• Tensor-based HMM
  • Seq2Prob: Convert observation sequences to co-occurrence matrix/tensor (order 2/3, lag, smoothing)
  • HMM: HMM parameter estimation via matrix/tensor decomposition (solvers: symNMF, SVD, CP, TT)
• Data Generation
  • toyModel: Synthetic HMM data (simple, weather, leftright)

References

• HMM: Rabiner, L. R., "A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition", Proceedings of the IEEE, 77(2), 257-286, 1989
• NMF for HMM: Kuang, D. et al., "Symmetric Nonnegative Matrix Factorization for Graph Clustering", SDM, 2012
• Spectral/SVD: Hsu, D. et al., "A Spectral Algorithm for Learning Hidden Markov Models", Journal of Computer and System Sciences, 78(5), 1460-1480, 2012
• Tensor Decomposition: Anandkumar, A. et al., "Tensor Decompositions for Learning Latent Variable Models", JMLR, 15, 2773-2832, 2014
• Beta-divergence (MU rule): Fevotte, C. and Idier, J., "Algorithms for Nonnegative Matrix Factorization with the Beta-Divergence", Neural Computation, 23(9), 2011
• Beta-divergence (convergence): Nakano, M. et al., "Convergence-guaranteed Multiplicative Algorithms for Nonnegative Matrix Factorization with Beta-divergence", IEEE MLSP, 283-288, 2010

Contributing

If you have suggestions for how hmmTensor could be improved, or want to report a bug, open an issue! We'd love all and any contributions.

For more, check out the Contributing Guide.

Authors

• Koki Tsuyuzaki