Lyapunov spectrum of random neural networks
In the authors' words
The Lyapunov spectrum of a nonlinear recurrent neural network with random asymmetric couplings is calculated in the limit . The calculation is based on a finite- identity that expresses the cumulative distribution of Lyapunov exponents as a response function of the tangent-space dynamics, with the tangent trajectory selected by a minimum-norm condition rather than by an initial condition. A cavity method then determines this response function at large through a self-consistent single-site problem. This result establishes that the chaos in this network is extensive and gives access to diffeomorphism-invariant properties of the dynamics. This work was done in collaboration with the AI models GPT-6 Astra and Claude Opus 5.5.
Appeared: Friday, October 9. arXiv. Preprint, not yet peer-reviewed.
Authors' comment: 36 pages, 4 figures