2025-11-17 | | Total: 2
Continuous-Time Recurrent Neural Networks (CTRNNs) have been widely used for their capacity to model complex temporal behaviour. However, their internal dynamics often remain difficult to interpret. In this paper, we propose a new class of CTRNNs based on Hopfield-like associative memories with asymmetric couplings. This model combines the expressive power of associative memories with a tractable mathematical formalism to characterize fluctuations in nonequilibrium dynamics. We show that this mathematical description allows us to directly compute the evolution of its macroscopic observables (the encoded features), as well as the instantaneous entropy and entropy dissipation of the system, thereby offering a bridge between dynamical systems descriptions of low-dimensional observables and the statistical mechanics of large nonequilibrium networks. Our results suggest that these nonequilibrium associative CTRNNs can serve as more interpretable models for complex sequence-encoding networks.
We hybridize swap Monte Carlo and geometric relaxation to determine the intrinsic compositional structure (CS) of lead magnesium niobate (PMN) relaxor. We verify the stability of a Nb-rich sublattice in PMN, as prescribed by the prevailing random-site model. But ions in the complementary sublattice are not randomly mixed. The majority of Nb ions collapse into a dominant Nb cluster which percolates the lattice with a mesh-like geometry that prevents large space charges. Subsequent molecular dynamics simulations predict a pair distribution function agreeing with neutron scattering experiments. Analysis of dipolar structures in the Nb cluster sheds light on the unique dielectric property of PMN.