GitHub - akpradhn-zz/tf_hmm: Hidden Markov Models (Gaussian emissions) in Python 3 using Tensorflow (forward-backward and expectation-maximization algorithms)

# imports
from tf_hmm import HiddenMarkovModel
from toy_dataset import toy_dataset
import tensorflow as tf
import time
%matplotlib inline

dataset = toy_dataset(1000,32)

hmm = HiddenMarkovModel(2,2,reports=True)

hmm.expectation_maximization(dataset,max_steps=1000)

the computation graph has been recreated in 2.5 sec
the training process has been converged in 13 steps in 2.4 sec

hmm.plot()

hmm.mu

array([[ 1.80040165, -2.15167103],
       [-1.29833214,  1.11217551]])

hmm.cov

array([[[ 1.6233517 , -0.14151129],
        [-0.14151129,  1.07758101]],

       [[ 2.29823099,  0.32517588],
        [ 0.32517588,  1.28161531]]])

hmm.p0

array([ 0.15488769,  0.84511231])

hmm.tp

array([[ 0.52553154,  0.47446846],
       [ 0.19438977,  0.80561023]])

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
.idea		.idea
Dockerfile		Dockerfile
README.md		README.md
hmm_clustering.ipynb		hmm_clustering.ipynb
jupyter_notebook_config.py		jupyter_notebook_config.py
notebook.ipynb		notebook.ipynb
output_4_0.png		output_4_0.png
run_jupyter.sh		run_jupyter.sh
run_tensorboard.sh		run_tensorboard.sh
test.py		test.py
tf_hmm.py		tf_hmm.py
toy_dataset.py		toy_dataset.py
toy_dataset_2.py		toy_dataset_2.py

Provide feedback