Lasagne/nolearn autoencoder - how to get hidden layer output?

Question

I've trained an autoencoder using lasagne/nolearn. Suppose the network layers are [500, 100, 100, 500]. I've trained the neural net like so:

net.fit(X, X)

I want to do something like the following:

net.predict(X, layer=2)

so I'll get the suppressed representation of my data. So, if my initial data have a shape [10000, 500], the resulting data will be [10000, 100].

I searched but could not find how to do that. Is it possible with lasagne/nolearn?

Answer 1

Looks like the answer is here in the documentation: http://lasagne.readthedocs.org/en/latest/user/layers.html#propagating-data-through-layers

Here are the relevant parts:

To compute the output of a network, you should instead call lasagne.layers.get_output() on it. This will traverse the network graph.

You can call this function with the layer you want to compute the output expression for:

>>> y = lasagne.layers.get_output(l_out)

In that case, a Theano expression will be returned that represents the output in function of the input variables associated with the lasagne.layers.InputLayer instance (or instances) in the network

...

You can also specify a Theano expression to use as input as a second argument to lasagne.layers.get_output():

>>> x = T.matrix('x')
>>> y = lasagne.layers.get_output(l_out, x)
>>> f = theano.function([x], y)

Assuming net is of type nolearn.lasagne.NeuralNet it looks like you can get access to the the underlying layer objects with net.get_all_layers(). I don't see it in the documentation but it's here on line 592.