Transfer Learning for Shallow NN?

Is it possible to use transfer learning on neural networks with one hidden layer? If not, what methods are available to use already trained networks on a different feature space?