When class membership probabilities are not provided by the user, the should be computed from the training data.

senarvi / theanolm

TheanoLM is a recurrent neural network language modeling tool implemented using Theano

Apache License 2.0

81 stars 29 forks source link

When class membership probabilities are not provided by the user, the should be computed from the training data. #6

Closed senarvi closed 8 years ago

senarvi commented 8 years ago

When --dictionary-format classes is used, i.e. the dictionary does not contain class membership probabilities, compute them from the word unigram probabilities in the training data.

senarvi commented 8 years ago

This has been implemented.