1: \begin{abstract}

2: 

3: Generative adversarial networks (GANs) are a class of machine-learning models that use adversarial training to generate new samples with the same (potentially very complex) statistics as the training samples. 

4: One major form of training failure, known as mode collapse, involves the generator failing to reproduce the full diversity of modes in the target probability distribution. 

5: %We present an effective model of GAN training which captures the learning dynamics of a wide generator neural networks by replacing the generator with a collection of $N$ particles coupled by a static NTK. We argue for a form of universality, in which these wide neural networks may be characterized by two numbers. 

6: Here, we present an effective model of GAN training, which captures the learning dynamics by replacing the generator neural network with a collection of particles in the output space; particles are coupled by a universal kernel valid for certain wide neural networks and high-dimensional inputs.

7: The generality of our simplified model allows us to study the conditions under which mode collapse occurs. 

8: Indeed, experiments which vary the effective kernel of the generator reveal a mode collapse transition, the shape of which can be related to the type of discriminator through the frequency principle.

9: Further, we find that gradient regularizers of intermediate strengths can optimally yield convergence through critical damping of the generator dynamics. Our effective GAN model thus provides an interpretable physical framework for understanding and improving adversarial training.

10: 

11: % power of deep generative models

12: % what is GAN: G & D => saddle point of loss landscape, dynamics is hidden

13: % one major form of failure: MC 

14: % effective model 

15: % physical analogy & interpretation allow understanding of the origin of MC

16: % provide a guide to resolve MC and better train GANs

17: 

18: \end{abstract}

19: