Search papers, labs, and topics across Lattice.
The paper introduces "drift-diffusion matching," a framework for training continuous-time RNNs with asymmetric connectivity to embed arbitrary stochastic dynamical systems within a low-dimensional latent space. This approach allows RNNs to faithfully represent the drift and diffusion components of stochastic differential equations, including nonlinear and nonequilibrium dynamics. The authors demonstrate the framework's utility by constructing RNNs that model associative and sequential memory through transient exploration of attractors, and they provide a method for decomposing the RNN dynamics based on connectivity asymmetry and time-irreversibility.
Unlock richer neural dynamics: Asymmetric RNNs, trained with drift-diffusion matching, can now faithfully embed complex stochastic systems, including chaotic attractors, opening doors to more biologically plausible and computationally powerful models.
Recurrent neural networks (RNNs) provide a theoretical framework for understanding computation in biological neural circuits, yet classical results, such as Hopfield's model of associative memory, rely on symmetric connectivity that restricts network dynamics to gradient-like flows. In contrast, biological networks support rich time-dependent behaviour facilitated by their asymmetry. Here we introduce a general framework, which we term drift-diffusion matching, for training continuous-time RNNs to represent arbitrary stochastic dynamical systems within a low-dimensional latent subspace. Allowing asymmetric connectivity, we show that RNNs can faithfully embed the drift and diffusion of a given stochastic differential equation, including nonlinear and nonequilibrium dynamics such as chaotic attractors. As an application, we construct RNN realisations of stochastic systems that transiently explore various attractors through both input-driven switching and autonomous transitions driven by nonequilibrium currents, which we interpret as models of associative and sequential (episodic) memory. To elucidate how these dynamics are encoded in the network, we introduce decompositions of the RNN based on its asymmetric connectivity and its time-irreversibility. Our results extend attractor neural network theory beyond equilibrium, showing that asymmetric neural populations can implement a broad class of dynamical computations within low-dimensional manifolds, unifying ideas from associative memory, nonequilibrium statistical mechanics, and neural computation.
