Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
(SystemTrajectory, LossGradient) -> ParameterGradients
Solve an adjoint differential equation backward in time to compute the gradients of a loss function with respect to system parameters.
Problem it solves
Computing gradients for parameter estimation or neural differential equations with forward-mode automatic differentiation scales poorly when there are many parameters.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.