apaam/mlpack
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Handle cases where scales are 0 correctly.

Browse Source
This commit is contained in:
Ryan Curtin
2014-01-14 20:50:23 +00:00
parent 4c63690e40
commit c4d0ef41bc
1 changed files with 10 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/mlpack/methods/hmm/hmm_impl.hpp
+10 -6
View File
@@ -430,9 +430,10 @@ void HMM<Distribution>::Forward(const arma::mat& dataSeq,
forwardProb(state, 0) = transition(state, 0) *
emission[state].Probability(dataSeq.unsafe_col(0));
// Then normalize the column.
// Then normalize the column, but only if the scale is not 0.
scales[0] = accu(forwardProb.col(0));
forwardProb.col(0) /= scales[0];
if (scales[0] != 0.0)
forwardProb.col(0) /= scales[0];
// Now compute the probabilities for each successive observation.
for (size_t t = 1; t < dataSeq.n_cols; t++)
@@ -447,9 +448,10 @@ void HMM<Distribution>::Forward(const arma::mat& dataSeq,
emission[j].Probability(dataSeq.unsafe_col(t));
}
// Normalize probability.
// Normalize probability, but only if the scale is not 0.
scales[t] = accu(forwardProb.col(t));
forwardProb.col(t) /= scales[t];
if (scales[t] != 0.0)
forwardProb.col(t) /= scales[t];
}
}
@@ -478,8 +480,10 @@ void HMM<Distribution>::Backward(const arma::mat& dataSeq,
backwardProb(j, t) += transition(state, j) * backwardProb(state, t + 1)
* emission[state].Probability(dataSeq.unsafe_col(t + 1));
// Normalize by the weights from the forward algorithm.
backwardProb(j, t) /= scales[t + 1];
// Normalize by the weights from the forward algorithm, if the scale is
// not 0.
if (scales[t + 1] != 0.0)
backwardProb(j, t) /= scales[t + 1];
}
}
}