A basic question in learning theory is to identify if two distributions are identical when we have access only to examples sampled from the distributions. This basic task is considered, for example, in the context of Generative Adversarial Networks (GANs), where a discriminator is trained to distinguish between a real-life distribution and a synthetic distribution. Classically, we use a hypothesis class H and claim that the two distributions are distinct if for some h ? H the expected value on the two distributions is (significantly) different. Our starting point is the following fundamental problem: "is having the hypothesis dependent on more than a single random example beneficial". To address this challenge we define k-ary based discriminators, which have a family of Boolean k-ary functions G. Each function g ? G naturally defines a hyper-graph, indicating whether a given hyper-edge exists. A function g ? G distinguishes between two distributions, if the expected value of g, on a k-tuple of i.i.d examples, on the two distributions is (significantly) different. We study the expressiveness of families of k-ary functions, compared to the classical hypothesis class H, which is k = 1. We show a separation in expressiveness of k + 1-ary versus k-ary functions. This demonstrate the great benefit of having k = 2 as distinguishers. For k = 2 we introduce a notion similar to the VC-dimension, and show that it controls the sample complexity. We proceed and provide upper and lower bounds as a function of our extended notion of VC-dimension.
|Journal||Advances in Neural Information Processing Systems|
|State||Published - 2019|
|Event||33rd Annual Conference on Neural Information Processing Systems, NeurIPS 2019 - Vancouver, Canada|
Duration: 8 Dec 2019 → 14 Dec 2019