I have traine an HMM for a binary classifications. know i wand to compute the probabilitiy of the most likely sequence. theprobability and LnProbabilities are always Infinite or Nan. I have to mention that the Osbservation sequence is 1100 long.
but in othet parts, as example whern i want to calculate alpha in forward procedure just the first 7 times i have a probability. from 7 till end this is also Nan.
the data dest are Observation vectors all in o/1 and HMM have 3 states
this is how its look like:
{1.0490671718016382E42,0.0,0.0},
{8.20896030464108E83,0.0,0.0},
{6.4235190171325645E125,0.0,0.0},
{5.026409561285822E167,0.0,0.0},
{3.9331701222336614E209,0.0,0.0},
{3.0777092518648974E251,0.0,0.0},
{2.4083103310149818E293,0.0,0.0},
{Infinity,0.0,0.0},
{Infinity,NaN,NaN},
{NaN,NaN,NaN},
....
,{NaN,NaN,NaN}}]
prob at iteration 0: Infinity
every single point of view would be helpfull
Thanks
I have traine an HMM for a binary classifications. know i wand to compute the probabilitiy of the most likely sequence. theprobability and LnProbabilities are always Infinite or Nan. I have to mention that the Osbservation sequence is 1100 long. but in othet parts, as example whern i want to calculate alpha in forward procedure just the first 7 times i have a probability. from 7 till end this is also Nan. the data dest are Observation vectors all in o/1 and HMM have 3 states this is how its look like: {1.0490671718016382E42,0.0,0.0}, {8.20896030464108E83,0.0,0.0}, {6.4235190171325645E125,0.0,0.0}, {5.026409561285822E167,0.0,0.0}, {3.9331701222336614E209,0.0,0.0}, {3.0777092518648974E251,0.0,0.0}, {2.4083103310149818E293,0.0,0.0}, {Infinity,0.0,0.0}, {Infinity,NaN,NaN}, {NaN,NaN,NaN}, .... ,{NaN,NaN,NaN}}] prob at iteration 0: Infinity
every single point of view would be helpfull Thanks