如何理解tensorflow中的“viterbi_decode”

Question

传统的隐马尔可夫模型( HMM )中的维特比算法有一个起始概率矩阵(viterbi algorithm wiki)，而tensorflow中viterbi_decode的参数只需要转移概率矩阵和发射概率矩阵。如何理解它？

def viterbi_decode(score, transition_params):
  """Decode the highest scoring sequence of tags outside of 
  TensorFlow.

  This should only be used at test time.

  Args:
    score: A [seq_len, num_tags] matrix of unary potentials.
    transition_params: A [num_tags, num_tags] matrix of binary potentials.

  Returns:
    viterbi: A [seq_len] list of integers containing the highest scoring tag
    indicies.
    viterbi_score: A float containing the score for the Viterbi 
    sequence.
  """

Answer 1

Tensorflow中的viterbi算法不需要初始概率矩阵，因为它通过为所有状态提供零概率来开始解码。

这意味着它从状态0开始。

您可以查看实现here。

Answer 2

我已经用tensorflow创建了关于维特比算法的完整详细的教程，你可以在这里看看：

假设您的数据如下所示：

# logits :       A [batch_size, max_seq_len, num_tags] tensor of unary potentials to use as input to the CRF layer.

# labels_a :     A [batch_size, max_seq_len] matrix of tag indices for which we compute the log-likelihood.

# sequence_len : A [batch_size] vector of true sequence lengths.

然后

log_likelihood , transition_params = tf.contrib.crf.crf_log_likelihood(logits,labels_a,sequence_len)

#return of crf log_likelihood function

# log_likelihood: A scalar containing the log-likelihood of the given sequence of tag indices.

# transition_params: A [num_tags, num_tags] transition matrix. 

# This is either provided by the caller or created in this function.

现在我们可以计算viterbi分数了：

# score: A [seq_len, num_tags] matrix of unary potentials.
# transition_params: A [num_tags, num_tags] matrix of binary potentials.

Notebook Link