A hack for less verbose initialization of nolearn.lasagne neural nets.

To use this, import layers from lasagnemould instead of lasagne and initialize your layers like this:

from lasagnemould import layers

mylayers = [
    layers.InputLayer(shape=(None, 784)),
    layers.DenseLayer(100),
    layers.DenseLayer(10, nonlinearity=softmax)
]

The advantage of this is that you can directly instantiate the layers -- including the use of *args -- instead of using factories and without the need to specify the incoming keyword.

See here for how this initialization differs from the default initialization.