Skip to content

Latest commit

 

History

History
44 lines (27 loc) · 1.5 KB

BestSequenceLayer.md

File metadata and controls

44 lines (27 loc) · 1.5 KB

CBestSequenceLayer Class

This class implements a layer that finds the optimal class sequence using the output of the CCrfLayer.

To use it during training as well, turn on the SetBestPrevClassEnabled setting in the corresponding CCrfLayer.

Settings

The layer has no settings.

Trainable parameters

The layer has no trainable parameters.

Inputs

The layer has two inputs:

The first input accepts a blob with int data that contains the optimal class sequences (the first output of the CCrfLayer). The dimensions are:

  • BatchLength is the sequence length
  • BatchWidth is the number of sequences in the set
  • Channels is equal to the number of classes
  • the rest of the dimensions are equal to 1

The second input accepts a blob with float data that contains non-normalized logarithm of optimal class sequences probabilities (i.e. the second output of the CCrfLayer). It has the same dimensions as the first input.

Outputs

The single output is a blob with int data that contains the optimal class sequences. The blob dimensions are:

  • BatchLength is equal to the inputs' BatchLength
  • BatchWidth is equal to the inputs' BatchWidth
  • the other dimensions are equal to 1