Not sure I know what Neural network is either. Is that Nik software?
Well, machine learning doesn't fall within my area of software expertise (what little of it survives almost a dozen years after my retirement) and I suspect a purist would find fault with the technical accuracy of my explanation, so take this with a grain of salt.
A convolutional neural network is a type of application software that is designed to iteratively discover patterns in a large "training set" of inputs—in this case, images—which can be used to identify similar attributes in subsequent inputs. After evaluating many examples of the same general type during a "machine-learning" phase, the neural network will be able to determine whether subsequent inputs are members of that type. So if you feed a neural network a large number of pictures of monkeys, it will eventually "learn," or infer the existence of, certain attributes in those images that represent the essential visual elements of monkeyness. Then when given a new image to evaluate, it will be able to distinguish between
monkey and
not-monkey.
The way I think of it is that I'm using the neural network backwards. I feed it a photograph and the neural network finds attributes in the image that match patterns it has been trained to recognize—e.g., what a woodcut looks like—then modifies those attributes to make the image look more like the patterns it has been trained on.
I don't recall for certain which online tool I used for this particular image, but it might well have been
Deep Dream Generator, an offshoot of a Google "artificial intelligence" research project.
