Several figures from Abhinav’s paper “Neural Closure Models for Dynamical Systems” have been featured on the cover of the August 2021 issue of the Proceedings of the Royal Society A. Congratulations to Abhinav!
The background is a spatio-temporal zooplankton field, simulated using a complex nonlinear 5-component 1-D physical-biogeochemical model. Seasonal variability is forced through the surface photosynthetically-available radiation and mixed layer depth, each of which vary in time. The 5-component model is one of the dynamical systems used to illustrate our novel neural closure modeling. Overlaid on the zooplankton field is the graphical representation of the time-discretized distributed neural delay differential equation (Distributed-nDDE). The blocks labeled DNN and the integral symbol represent any deep neural-network architecture and time-integration scheme.
