hmm-tool 0.0.4

A fast Python implementation of HMM.

A fast Python implementation of HMM

Installation

To install:

$ pip install hmm-tool

Quickstart

from hmmtool import HMM
import numpy as np

### 3 states, 6 observations:
### states values:0/1/2, observations values:0/1/2/3/4/5
hmm = HMM(3,6)

#train
s1 = np.random.randint(6,size = 60)
s2 = np.random.randint(6,size = 40)
hmm.add_data([s1,s2])
hmm.train(maxStep=50,delta=0.001)

#get params
print(hmm.pi)
print(hmm.A)
print(hmm.B)

#predict
#random data
s3 = np.random.randint(6,size = 10)
s4 = np.random.randint(6,size = 10)

#multi inputs:[[o1,o2,o3,...,ot1],[o1,o2,o3,...,ot2]]
#output: [prob1, prob2]
print(hmm.estimate_prob([s3,s4]))
#output: [(prob1, [s1,s2,s3,...,st1]), (prob2, [s1,s2,s3,...,st2])]
print(hmm.decode([s3,s4]))