Interpretable statistical representations of neural population dynamics and geometry
arXiv:2304.03376 · doi:10.1038/s41592-024-02582-2
Abstract
The dynamics of neuron populations commonly evolve on low-dimensional manifolds. Thus, we need methods that learn the dynamical processes over neural manifolds to infer interpretable and consistent latent representations. We introduce a representation learning method, MARBLE, that decomposes on-manifold dynamics into local flow fields and maps them into a common latent space using unsupervised geometric deep learning. In simulated non-linear dynamical systems, recurrent neural networks, and experimental single-neuron recordings from primates and rodents, we discover emergent low-dimensional latent representations that parametrise high-dimensional neural dynamics during gain modulation, decision-making, and changes in the internal state. These representations are consistent across neural networks and animals, enabling the robust comparison of cognitive computations. Extensive benchmarking demonstrates state-of-the-art within- and across-animal decoding accuracy of MARBLE compared with current representation learning approaches, with minimal user input. Our results suggest that manifold structure provides a powerful inductive bias to develop powerful decoding algorithms and assimilate data across experiments.
References in corpus (8)
- Deep learning for universal linear embeddings of nonlinear dynamics
- Chaos as an Intermittently Forced Linear System
- Linking connectivity, dynamics and computations in low-rank recurrent neural networks
- Learnable latent embeddings for joint behavioral and neural analysis
- Data-driven discovery of intrinsic dynamics
- Unfolding the multiscale structure of networks with dynamical Ollivier-Ricci curvature
- Topological metric detects hidden order in disordered media
- Implicit Gaussian process representation of vector fields over arbitrary latent manifolds