Publisher DOI: | 10.1103/PhysRevE.111.054208 | Title: | Weight dynamics of learning networks | Language: | English | Authors: | Sharafi, Nahal Martin, Christoph Hallerberg, Sarah |
Keywords: | Dynamical systems; Artificial neural networks; Chaos & nonlinear dynamics | Issue Date: | 12-May-2025 | Publisher: | American Physical Society | Journal or Series Name: | Physical review / publ. by The American Institute of Physics. E | Volume: | 111 | Issue: | 5 | Startpage: | 054208-1 | Endpage: | 054208-14 | Abstract: | Neural networks have become a widely adopted tool for tackling a variety of problems in machine learning and artificial intelligence. In this contribution, we use the mathematical framework of local stability analysis to gain a deeper understanding of the learning dynamics of feedforward neural networks. We derive equations for the tangent operator of the learning dynamics of three-layer networks learning regression tasks. The results are valid for an arbitrary number of nodes and arbitrary choices of activation functions. Applying the results to a network learning a regression task, we investigate numerically how stability indicators relate to the final training loss. Although the specific results vary with different choices of initial conditions and activation functions, we demonstrate that it is possible to predict the final training loss by monitoring finite-time Lyapunov exponents during the training process. |
URI: | https://hdl.handle.net/20.500.12738/17863 | ISSN: | 2470-0053 | Review status: | This version was peer reviewed (peer review) | Institute: | Department Maschinenbau und Produktion Fakultät Technik und Informatik |
Type: | Article | Additional note: | article number: 054208 |
Appears in Collections: | Publications without full text |
Show full item record
Add Files to Item
Note about this record
Export
This item is licensed under a Creative Commons License