When does Matlab consider a Neural Network deep?

I want to use a feedforward NN (FFNN) with at least two hidden layers. This is no problem with the feedforwardnet function. I also can train the net with the train function, where it says "Train shallow neural network". In most literature I read, a shallow NN is defiend as a NN with only one hidden layer and a deep NN is one with more than one hidden layer. For deep NNs Mathworks recomends trainNetwork as a train function. But I was not able to train my FFNN with this function.

So where does Mathworks draw the line between shallow and deep NNs? Or is there a way to train my FFNN with trainNetwork?

[Annette Mai]

I can delcare a FFNN with two hidden layers like this:

a = [1 2 3 4 5 6 7 8 9 10 7 8 9 10 11 12 13 14 15 16];
b = [7 8 9 10 11 12 13 14 15 16 1 2 3 4 5 6 7 8 9 10];
c = a.*b;

FFnet = feedforwardnet([3,4]);
FFnet = train(FFnet, [a;b], c)

[Walter Roberson]

"Deep" for Neural Network seems to be much the same as "deep" for philosophy:

1. It claims to encompass pretty much everything
2. Hardly anybody really understands the logic behind it

Answer 1

[Rohit Pappu]

trainNetwork works with Deep Learning Layers. Additional Information can be found here

[Annette Mai]

So for Mathworks a feedfoward net with two hidden layers is not a deep neural net?!

[Annette Mai]

I want to map a nonlinear equation with a NN. I´ve got data like in my other exampel a and b as Inputs and c as the resulting output. How could I load this data into the Deep Network Designer? All the examples seem to be about image processing.