hmmlearn_gaussian_hmm_manual_params_sample_generation.py

1import numpy as np
2from hmmlearn import hmm
3
4# Prepare parameters for a 3-state Gaussian HMM
5# Initial state probabilities
6startprob = np.array([0.6, 0.3, 0.1])
7# State transition probabilities
8transmat = np.array([[0.7, 0.2, 0.1],
9                      [0.3, 0.5, 0.2],
10                      [0.3, 0.3, 0.4]])
11# The means of each component
12means = np.array([[0.0,  0.0],
13                  [10.0, 10.0],
14                  [20.0, 20.0]])
15# The covariance of each component
16covars = np.tile(np.identity(2), (3, 1, 1))
17
18# Build the model
19model = hmm.GaussianHMM(n_components=3, covariance_type="full")
20
21# Instead of fitting it from data, we set the parameters manually
22model.startprob_ = startprob
23model.transmat_ = transmat
24model.means_ = means
25model.covars_ = covars
26
27# Generate samples
28X, Z = model.sample(100)
29
30print("Observations (X):")
31print(X[:5])
32print("States (Z):")
33print(Z[:5])