Making Deep learning Layer : 1D conv like [1 25] filter
显示 更早的评论
hello,
I am author of gCNN method,
i used matconvnet toolbox, but i want to adapt matlab's own toolbox,
but i have had many trouble with making new class, if someone help me, i make a great speed of development...
- i found, Define Custom Deep Learning Layer - "https://kr.mathworks.com/help/deeplearning/ug/define-custom-deep-learning-layer.html"
- so, i do make a new convolutionLayer using already exist class like Convolution2D.m
- my list of class below
* Class location : /Applications/MATLAB_R2019a.app/toolbox/nnet/cnn/+nnet/+internal/+cnn/+layer/
1) geometryConvolution2D.m
2) geometryMaxPooling2D.m
* Class location : /Applications/MATLAB_R2019a.app/toolbox/nnet/cnn/+nnet/+internal/+cnn/+layer/+util
3) geometryConvolution2DGPUStrategy.m
4) geometryConvolution2DHostStridedConvStrategy.m
1) geometryConvolution2D.m :: doForward (in geometryConvolution2D.m > forward > forward normal)
function [Z, memory] = predictNormal( this, X )
[weights0,bias0] = this.QuantizationMethod.remapped(this.Weights.Value,this.Bias.Value);
% Data preprocessing
X_ = mnet_make25dot_interpolation(X); % my owen function to sampling data N x 1 to N x 25 // conv 1D filter 1x25 then, get N x 1 again
Z = this.doForward(X_,weights0,bias0);
%
memory = [];
end
my bottleneck is here...
% Set Hyper-parameters
this.FilterSize = [1, filterSize(2) ]; % use same code :: convolution2dLayer_sb(25, 20, 'Name', 'Gconv_1') % custom Layer
but it does not work... error to check function in "trainNetwork.m"
% Infer parameters of layers
isaDAG = iHaveDAGNetwork(layersOrGraph);
analyzedLayers = iInferParameters(layersOrGraph); % !!!!!!!!!here!!!!!!!!!
please help me
1 个评论
Dheeraj Singh
2019-8-5
Could you provide the code files and the error message that you received...
回答（0 个）
类别
在 帮助中心 和 File Exchange 中查找有关 Deep Learning Toolbox 的更多信息
2019-7-10
2019-8-5
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
Translated by 
