ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2302.11055
79
73

SGD learning on neural networks: leap complexity and saddle-to-saddle dynamics

21 February 2023
Emmanuel Abbe
Enric Boix-Adserà
Theodor Misiakiewicz
    FedML
    MLT
ArXivPDFHTML
Abstract

We investigate the time complexity of SGD learning on fully-connected neural networks with isotropic data. We put forward a complexity measure -- the leap -- which measures how "hierarchical" target functions are. For ddd-dimensional uniform Boolean or isotropic Gaussian data, our main conjecture states that the time complexity to learn a function fff with low-dimensional support is Θ~(dmax⁡(Leap(f),2))\tilde\Theta (d^{\max(\mathrm{Leap}(f),2)})Θ~(dmax(Leap(f),2)). We prove a version of this conjecture for a class of functions on Gaussian isotropic data and 2-layer neural networks, under additional technical assumptions on how SGD is run. We show that the training sequentially learns the function support with a saddle-to-saddle dynamic. Our result departs from [Abbe et al. 2022] by going beyond leap 1 (merged-staircase functions), and by going beyond the mean-field and gradient flow approximations that prohibit the full complexity control obtained here. Finally, we note that this gives an SGD complexity for the full training trajectory that matches that of Correlational Statistical Query (CSQ) lower-bounds.

View on arXiv
Comments on this paper