Table of Contents  ▸  List of Commands  ▸  Neural Networks  ▸  nn_conv

◀  nn_columnpatch2d    |    nn_conv2d  ▶

nn_conv

Arguments:

• out,in,nb_channels>0,_kernel_width>0,_kernel_height>0,_kernel_depth>0,_stride_x>0,_stride_y>0,_stride_z>0,_dilation_x,_dilation_y,_dilation_z,_shrink_x>=0,_shrink_y>=0_,_shrink_z>=0,boundary_conditions,_learning_mode,_weight_decay>=0,_initialization

Description:

Add a conv layer (generic 1D/2D/3D convolutional layer) to the current network.

boundary_conditions can be { 0:Dirichlet | 1:Neumann | 2:Periodic | 3:Mirror }.
learning_mode can be { 0:No learning | 1:Weights only | 2:Biases only | 3:Weights+biases }.
initialization can be { 0:Zero | 1:identity | 2:Lecun-initialization | 3:He-initialization | 4:Residual He-initialization }.

Default values:

stride_x=stride_y=stride_z=1, dilation_x=dilation_y=dilation_z=1, shrink_x=shrink_y=shrink_z=0, boundary_conditions=1, learning_mode=3, weight_decay=0 and initialization=3.

G'MIC is an open-source software distributed under the CeCILL free software licenses (LGPL-like and/or
GPL-compatible). Copyrights (C) Since July 2008, David Tschumperlé - GREYC UMR CNRS 6072, Image Team.