Adds new layer to the net

id = net.addLayer(name, layerType, params)

Input

• name unique name of the adding layer.
• layerType typename of the adding layer (type must be registered).
• params parameters which will be used to initialize the creating layer. Scalar structure with the following fields:
  • dict name-value dictionary as struct, values are scalar values (or arrays) of one of the following type: double, integer, or string.
  • blobs List of learned parameters stored as blobs.
  • name Name of the layer instance (optional, can be used internal purposes).
  • type Type name which was used for creating layer by layer factory (optional).

Output

• id unique identifier of created layer, or -1 if a failure will happen.

A LayerParams provides all data needed to initialize layer. It includes dictionary with scalar params (params.dict struct), blob params params.blobs and optional meta information params.name and params.type of layer instance.

Built-in layers listed below partially reproduce functionality of corresponding Caffe and Torch7 layers. In partuclar, the following layers and Caffe importer were tested to reproduce Caffe functionality:

• Convolution
• Deconvolution
• Pooling
• InnerProduct
• TanH, ReLU, Sigmoid, BNLL, Power, AbsVal
• Softmax
• Reshape, Flatten, Slice, Split
• LRN
• MVN
• Dropout (since it does nothing on forward pass)