Package: pinnr 0.1.1

pinnr: Physics-Informed Neural Networks in Native R with 'torch'

Fits physics-informed neural networks (PINNs) that solve ordinary and partial differential equations and estimate unknown model parameters from data. Networks are trained by penalising the equation residual, boundary and initial conditions, and misfit to observations, following Raissi, Perdikaris and Karniadakis (2019) <doi:10.1016/j.jcp.2018.10.045>. Everything runs in native R on top of 'torch', with no 'Python' dependency, and ordinary differential equation models written as 'deSolve'-style derivative functions can be used directly.

Authors:Samson Obasanmi [aut, cre]