English

Global Optimality of Elman-type RNN in the Mean-Field Regime

Machine Learning 2023-03-14 v1 Machine Learning

Abstract

We analyze Elman-type Recurrent Reural Networks (RNNs) and their training in the mean-field regime. Specifically, we show convergence of gradient descent training dynamics of the RNN to the corresponding mean-field formulation in the large width limit. We also show that the fixed points of the limiting infinite-width dynamics are globally optimal, under some assumptions on the initialization of the weights. Our results establish optimality for feature-learning with wide RNNs in the mean-field regime

Keywords

Cite

@article{arxiv.2303.06726,
  title  = {Global Optimality of Elman-type RNN in the Mean-Field Regime},
  author = {Andrea Agazzi and Jianfeng Lu and Sayan Mukherjee},
  journal= {arXiv preprint arXiv:2303.06726},
  year   = {2023}
}

Comments

31 pages, 2 figures