I'm having a problem in running NN training with
net = trainNetwork(trainingSet, layers, opts);
count = gpuDeviceCount --> 0 [I do not have a GPU. Matlab is 2016b version]
I also tried setting the value of nnet.internal.cnngpu.isGPUCompatible to false but it wasn't possible. (uneditable matlab builtin function)
I used activations earlier
trainingFeatures = activations(convnet, trainingSet, featureLayer, ...
'MiniBatchSize', 32,'OutputAs', 'columns');
But I'm not sure How I can use 'layers' in the same way with activations.. Anyhelp would be highly appreciated. s
You can try some deep learning examples without GPU in matlab deep learning toolbox examples, such as deep autoencoders.
You can qlso download some deep learning matlab toolboxes without GPU by google from the Internet.
Check this link
https://www.mathworks.com/discovery/deep-learning.html
For convolotion neural network you can use toolboxes such as matconnet, and caffe in matalb without gpu.
follow these links
http://www.vlfeat.org/matconvnet/
and
for
caffe
http://caffe.berkeleyvision.org/
https://www.youtube.com/watch?v=AN2uXGRvw9E
you can build and use without GPU and without
Hi,
There is nice video from youtube which shows you how to use deep learning to recognize objects
https://www.youtube.com/watch?v=-ENmRfKWjmo
try: https://www.mathworks.com/matlabcentral/fileexchange/60659-deep-learning-in-11-lines-of-matlab-code
MatConvNet also uses 'ExecutionEnvironment','cpu' option that is not supported by matlab r2016b. so how you resolved the issue? As i am facing with same problem
no way, until upgade it to matlab-2017 if using TrainNetwork command
If I use matlab 2017a but still find error and can't work.
What will be the solution?
@ Huma Chaudhry
in Training option, you keep ExecutionEnvironment','cpu' .
- Badges
- Science topic