Module
network
A module to build neural network from a description.
An example of description can be found in this file: Inference.lua
Info:
- edit Alexis BRENON alexis.brenon@imag.fr
Data Types
keyboard_arrow_up-
ConvolutionDescription
Description of a convolutionnal layer.
Fields:-
integer
n_filters
Number of filters to use -
DimenTable
field_size
Size of the receptive field -
DimenTable
stride
Size of the strides -
DimenTable
zero_padding
Size of the paddings
-
integer
-
DimenTable
-
NetworkDescription
Description of a neural networks.
Fields:-
table
input_size
Size of the input Tensor{d, w, h}
-
{ConvolutionDescription,...}
conv_layers
List of convolution layers -
{number,...}
fc_layers
List of width of nn.Linear fully connected layers -
string
nl_layer
Name of the transfert function to use -
table
output_size
Size of the output Tensor{w, h}
-
table
Public Methods
keyboard_arrow_up-
create_network ( args )
Build a network.
Parameters:-
NetworkDescription
args
Description of the network to build
- nn.Sequential Built network
-
NetworkDescription