Deep Neural Network Tranining

2 次查看（过去 30 天）
Veton Kepuska
Veton Kepuska 2019-12-14
编辑： Piyush Dubey 2019-12-17
Hi,
I am having trouble in running the modified DeepLearningSpeechRecognitionExample_all that utilzis all the data:
"bird",...
"cat",...
"dog",...
"down",...
"eight",...
"five",...
"four",...
"happy",...
"house",...
"left",...
"marvin",...
"nine",...
"one",...
"right",...
"seven",...
"sheila",...
"six",...
"stop",...
"three",...
"two",...
"zero",...
"bed",...
"go",...
"no",...
"off",...
"on",...
"tree",...
"up",...
"wow",...
and modifyed parametrs in training (double number of feature vectors) that is:
segmentDuration = 1;
frameDuration = 0.020; %0.025;
hopDuration = 0.005; %0.010; <---------------- doubleing the sigze of feature vectors
numBands = 40; %40;
as well as more layers of neural network and larger feature vector size.
numF = 40; %12;
layers = [
imageInputLayer(imageSize)
convolution2dLayer(3,numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
maxPooling2dLayer(3,'Stride',2,'Padding','same')
convolution2dLayer(3,2*numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
maxPooling2dLayer(3,'Stride',2,'Padding','same')
convolution2dLayer(3,3*numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
maxPooling2dLayer(3,'Stride',2,'Padding','same')
convolution2dLayer(3,4*numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
maxPooling2dLayer(3,'Stride',2,'Padding','same')
convolution2dLayer(3,5*numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
maxPooling2dLayer(3,'Stride',2,'Padding','same')
convolution2dLayer(3,4*numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
maxPooling2dLayer(3,'Stride',2,'Padding','same')
convolution2dLayer(3,3*numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
maxPooling2dLayer(3,'Stride',2,'Padding','same')
convolution2dLayer(3,2*numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
maxPooling2dLayer(3,'Stride',2,'Padding','same')
convolution2dLayer(3,numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
convolution2dLayer(3,numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
convolution2dLayer(3,numF,'Padding','same')
batchNormalizationLayer
reluLayer
%tanhLayer
dropoutLayer(dropoutProb)
fullyConnectedLayer(numClasses)
softmaxLayer
weightedClassificationLayer(classWeights)];
The code fials with this error:
...done
Training error: 1.2073%
Validation error: 3.8136%
Network size: 4772.4043 kB
Error using classify (line 149)
The length of GROUP must equal the number of rows in TRAINING.
Error in DeepLearningSpeechRecognitionExample_all (line 457)
[YPredicted,probs] = classify(1,x,"ExecutionEnvironment",'cpu');
I need a profesional help to figure out why am getting the length difference? Note that the reson why I am not able to track that problem down is due to the fact that this program requires at least a half-day to run in my computer before it failes.
Thank you
--Veton

请先登录，再回答此问题。

回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 AI for Signals 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by