MATLAB: How to resolve Error using trainNetwork (line 140) Can’t reload ‘/Applications/MATLAB_R2017b.app/bin/maci64/libmwgpu.dylib’

can't reload (mac)Deep Learning Toolbox

The script below worked yesterday, but not today after installing the Parallel Computing Toolbox!

The imageDataStore loaded fine; the Path is OK. I am using a MAC (OS Mavericks 10.9.5) with no GPU. I am using MATLAB R2017b.

>> flowerds = imageDatastore('Flowers','IncludeSubfolders',true,'LabelSource','foldernames');

>> [trainImgs,testImgs] = splitEachLabel(flowerds,0.6);

>> numClasses = numel(categories(flowerds.Labels));

>> net = alexnet;

>> prebuiltLayers = net.Layers(1:end-3);

>> layers(end-2) = fullyConnectedLayer(numClasses);

>> layers(end) = classificationLayer;

>> topts = trainingOptions('sgdm','InitialLearnRate',0.001);

>> [flowernet,info] = trainNetwork(trainImgs, layers, topts);

Error using trainNetwork (line 140)

Can't reload '/Applications/MATLAB_R2017b.app/bin/maci64/libmwgpu.dylib'

Caused by:

Can't reload '/Applications/MATLAB_R2017b.app/bin/maci64/libmwgpu.dylib'

Best Answer

Mac OS Mavericks is not supported for MATLAB R2017b, according to the following documentation pages:

road map

system requirements

The following workaround might be able to resolve the errors, but please be aware that there may be other functionalities that no longer work in an unsupported platform of a specific MATLAB release.

In the step in the code where you set the training options (the second-to-last line), please try manually/explicitly setting the "ExecutionEnvironment" to use the CPU. Here is the modified command to try:

>> topts = trainingOptions('sgdm','InitialLearnRate',0.001,'ExecutionEnvironment','cpu');

Related Solutions

MATLAB: Error using activations with GoogLeNet in R2017b

The pretrained networks "alexnet" and "googlenet" belong to different MATLAB classes: "alexnet" is a SeriesNetwork, while "googlenet" is a DAGNetwork (where layers do not have to be arranged as one single chain). For a DAGNetwork, the "activations" method is not fully supported yet – this functionality will be available in a future MATLAB release.

But as you found out, the "activations" method does work on a "googlenet" to some extent. It returns a _h_-by-_w_-by-_c_-by-_n_ array, where _h_, _w_, and _c_ are the height, width, and number of channels for the output of the chosen layer, and _n_ is the number of observations.

The workaround is to reshape the matrix to be in the form of (number of observations) x (number of features) as is needed for "fitcecoc" for feature extraction. Here (number of features) = h*w*c. You should be able to reshape the training and testing matrices as follows:

>> nfeatures = size(trainingFeatures, 1) * size(trainingFeatures, 2) * size(trainingFeatures, 3);
>> nobservations = size(trainingFeatures, 4);
>> trainingFeatures = reshape(trainingFeatures, nfeatures, nobservations);
>> trainingFeatures = trainingFeatures';

For your reference, here are the documentation pages for SeriesNetwork and DAGNetwork:

www.mathworks.com/help/nnet/ref/seriesnetwork.html

www.mathworks.com/help/nnet/ref/dagnetwork.html

MATLAB: How to programmatically replace a referenced model with a Subsystem in Simulink

Assuming that you have a 'parent.slx' model which references a model 'child.slx' (whose block name is 'Model' in the parent Simulink model), use either of the following options:

>> open_system('parent');

>> add_block('built-in/Subsystem',['parent/mySub']);

>> load_system('child');

>> Simulink.BlockDiagram.copyContentsToSubsystem('child',['parent/mySub']);

>> pos = get_param('parent/Model','position');

>> delete_block('parent/Model');

>> set_param('parent/mySub','position',pos);

>> open_system('parent');

>> load_system('child');

>> convert2subsys('child');

>> add_block('child/Subsystem','parent/mySub');

>> pos = get_param('parent/Model','position');

>> delete_block('parent/Model');

>> set_param('parent/mySub','position',pos);

>> close_system('child',0);

where 'convert2subsys' is defined at the following link:

https://uk.mathworks.com/help/simulink/slref/simulink.blockdiagram.createsubsystem.html#br5mgzk-6

Please refer to the attachment for further details.

Best Answer

Related Solutions

MATLAB: Error using activations with GoogLeNet in R2017b

MATLAB: How to programmatically replace a referenced model with a Subsystem in Simulink

Related Question