# Do GAN's come under Supervised Learning or Unsupervised Learning?

My guess is that they come under supervised learning, as we have labelled dataset of images, but I am not sure as there maybe other aspects in GANs which might come into play in the determination of the class of algorithms GAN falls under.

• stackoverflow.com/questions/44445778/…. I think you'll find all your answers here. Actually you have a very good question. – DuttaA Nov 10 '18 at 7:37
• Hey, thanks for both the comments and edit, comment is useful. – codecracker Nov 11 '18 at 22:56

The terms Supervised Learning and Unsupervised Learning predate the invention of the application of artificial networks to a generative and discriminative network pair, which was the first popular generative topology. The existence of labeling is the key distinction between the two. Even partial labeling indicates supervision, as odd as that jargon is, since the supervisor does no learning and the labels are constants.

Based on the original description of the discriminative network in a GAN, that it consumes examples without labels, GANs are a type of unsupervised learning. That fact does not eliminate the use of labels as part of an extension of the original design that generates based on some labeled element in the examples or the use of other labels to indicate the fitness of each image for a class into which generated images are expected to fall.

GANs are unsupervised machine learning algorithms. According to Wikipedia, unsupervised algorithms are :

Unsupervised learning is a branch of machine learning that learns from test data that has not been labelled, classified or categorized

In GAN networks, only training data is provided which is not labelled. The network generates candidates ( generative ) which are evaluated by the discriminator. The network slowly learns from the data given from a latent space.

Suppose, you want to create a GAN network which can make a Monet painting. You just need to feed it some Monet paintings. Here, you are not interested in classifying the painting, but copying/mimicking it.

Hence, there is no need for labels here which makes GAN an unsupervised machine learning algorithm.

• when you feed some monet paintings, arent you providing labelled data with all labels as monet paintings? – codecracker Nov 10 '18 at 5:09
• That makes no sense. Every object, according to humans has some or the other class or label. Th main point here is that we are making no classification. – Shubham Panchal Nov 10 '18 at 5:37
• Not all supervised machine learning is classification. The point is that there is no single correct response which we want to get and about which we tell the algorithm – Martin Thoma Nov 10 '18 at 11:28

neither. roughly, the hierarchy looks like the following:

               machine learning methodology
+
|
|
v
+-----------------------------------------------+
|                        |                      |
|                        |                      |
v                        v                      v
supervised              unsupervised           reinforcement

• Can you extend your answer a bit? Where does GAN fit into this, and why? – Oliver Mason Feb 10 at 20:22