https://github.com/nhchristianson/poissonhmm

A Python library for working with and training Hidden Markov Models with Poisson emissions.
https://github.com/nhchristianson/poissonhmm

hmm hmm-model poisson poisson-distribution poisson-emissions python python-library training-hmms

Last synced: 7 months ago
JSON representation

A Python library for working with and training Hidden Markov Models with Poisson emissions.

• Host: GitHub
• URL: https://github.com/nhchristianson/poissonhmm
• Owner: nhchristianson
• License: mit
• Created: 2017-08-05T01:29:36.000Z (almost 8 years ago)
• Default Branch: master
• Last Pushed: 2017-08-14T14:08:33.000Z (almost 8 years ago)
• Last Synced: 2023-10-20T22:03:29.753Z (over 1 year ago)
• Topics: hmm, hmm-model, poisson, poisson-distribution, poisson-emissions, python, python-library, training-hmms
• Language: Python
• Homepage:
• Size: 5.86 KB
• Stars: 9
• Watchers: 2
• Forks: 4
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# PoissonHMM

A Python library for working with and training HMMs with Poisson emissions.

There are two classes in this library:

`PHMM` creates a typical HMM with Poisson emissions, where every sequence is assumed to have been generated with the same Poisson parameters - i.e., if the HMM has three states with Poisson means of 1.0, 3.0, and 4.5, every sequence will be generated using those parameters.

`PHMM_d` creates a Poisson-emitting HMM where sequences can be generated with different Poisson parameters. Hence, the parameters are formatted as a nested array, where each subarray is the set of emission parameters for a single sequence, and the length of the overall array is the number of observation sequences you'd like to train. This allows for the training of a PHMM such that the state transition matrix is trained over all observation sequneces, but state magnitudes can differ from sequence to sequence.