NetworkWrongType: First argument must be a SeriesNetwork, DAGNetwork, dlnetwork, or layerGraph.

Hi,

I am a beginner in the deep learning toolbox of Matlab, so I apologize if my question seems too simple. I created a simple network using the 'feedforwardnet' function (MyNet = feedforwardnet([10]);). After training the network, I want to export it for use in Fortran. The first step is to export the network using exportONNXNetwork . I did this as follows:

filename = "MyNet.onnx";
exportONNXNetwork(MyNet,filename)

And I got an error message saying:

Error using nnet.internal.cnn.onnx.exportONNXNetwork>iValidateNetwork (line 62)
First argument must be a SeriesNetwork, DAGNetwork, dlnetwork, or layerGraph.

Does that mean that feedforwardnet is not a valid type of network for export??

Answer 1

[Walter Roberson]

feedforward networks are "shallow" networks. Export to onyx only supports "deep" networks.

(I do not whether onyx networks support functionality equivalent to shallow networks. I do not know much about onyx networks)

[mary]

thank you

[mary]

@Walter Roberson@Divyank Could you recommend one of these networks that can replace Feedforwardnet ? My input and output data are not of image type and my objective is to do a regression rather than classification. Could you please tell me where I can read about the differences of these networks? and if possible, where can I find simple examples of their applications ?

[Walter Roberson]

See https://www.mathworks.com/help/deeplearning/deep-learning-with-time-series-sequences-and-text.html

Answer 2

[Divyank]

Hello @mary, that's correct, 'feedforwardnet' is not a supported network type for export to ONNX format using the exportONNXNetwork function. You need to use either a 'SeriesNetwork', 'DAGNetwork', 'dlnetwork', or 'layerGraph' type of network for export to ONNX. For more information please refer to the following documentation: Export Network to ONNX model format

[mary]

thank you