Deep dreaming refers to the generation of images that desired activations in a trained deep network.For example, given a network that is trained to recognize cats (among other things), a dreamed cat image can be synthesized by performing a Gradient descent optimization adjusting a random image such that when it is fed forward through the trained network, it produces the "this is a cat" output.The optimization resembles Backpropagation, however instead of adjusting the network weights,the weights are held fixed and the input is adjusted.