Finding separatrices of dynamical flows with Deep Koopman Eigenfunctions, Neurips 2025
A numerical method utilising neural networks and Koopman theory to find separatrices of dynamical systems.
When predict can also explain: few-shot prediction to select better neural latents
A prediction-based framework to evaluate the goodness of latent variable models of neural data. Few-shot prediction encourages ‘simplicity’, discouraging extraneous features in the latents.