MATLAB: Can I give mathworks a few tips on the development of the Deep Learning Toolbox

Deep Learning Toolboxdeeplearning

1,Hope the future versions support multi-tasking multi-label training, classified neural networks；

2,Enhance the versatility of the importCaffenetworks function, because many model imports in the model zoo do not come in, many layers do not support；

3，3D convolution, 3D pooling ,video classification needs to be implemented Thankyou! Forgive my English is poor

Best Answer

Hi Cui,

Thanks for your comment. We are actively working on the features you mentioned in your post.

We are continuously making the importCaffeNetworks more robust but providing support for all the imported layers is something we'd want to do on a case-by-case basis.

I'll connect with you separately to discuss more.

Thanks Shounak

Related Solutions

MATLAB: Weight values from “importCaffeNetwork”

The weights are stored in the layers themselves. For example:

>> network.Layers(5)
ans = 
  FullyConnectedLayer with properties:
          Name: 'ip1'
   Hyperparameters
     InputSize: 2880
    OutputSize: 10
   Learnable Parameters
       Weights: [10×2880 single]
          Bias: [10×1 single]

MATLAB: Deep learning toolboxes and Caffe

There are a lot of deep learning frameworks with MATLAB support. MathWork's own neural network toolbox has convolutional neural network support:

http://www.mathworks.com/help/nnet/convolutional-neural-networks.html

Here is a list of popular frameworks, look for the one's that list MATLAB as supported interfaces: https://developer.nvidia.com/deep-learning-frameworks

Best Answer

Related Solutions

MATLAB: Weight values from “importCaffeNetwork”

MATLAB: Deep learning toolboxes and Caffe

Related Question