search

zhongkaifu / RNNSharp

RNNSharp is a toolkit of deep recurrent neural network which is widely used for many different kinds of tasks, such as sequence labeling, sequence-to-sequence and so on. It's written by C# language and based on .NET framework 4.6 or above versions. RNNSharp supports many different types of networks, such as forward and bi-directional network, sequence-to-sequence network, and different types of layers, such as LSTM, Softmax, sampled Softmax and others.
BSD 3-Clause "New" or "Revised" License
285 stars 91 forks source link

Supported Features #21

Closed My-Khan closed 7 years ago

My-Khan commented 8 years ago

Hello Hope that you will be fine. Can you briefly explain impact of each feature type specially the embedding feature and runtime feature on the final results of RNN compared to the traditional CRF which use only one feature type e.g. the template feature. thanks in advance

zhongkaifu commented 8 years ago

Template feature is sparse feature. If current token hits some features, only these features values become non-zero, however, other features values are zero. Embedding feature is dense feature. each token can point to a vector and the value of this vector will be used as embedding feature. Runtime feature uses the output of previous tokens as features for current token.

My-Khan commented 8 years ago

Many thanks for details.