DiffEqFlux.jl – A Julia Library for Neural Differential Equations

We just released a new blog post showing how you can replicate the results of the recent Neural ODEs paper – and then some – using Flux and DifferentialEquations.jl. We’ve been combining ML and DiffEqs for a quite some time, just because it’s easy to do in Julia – all the tools exist already and we just need to combine them. Even so, the simple convenience functions we provide in DiffEqFlux make it particularly straightforward to replicate these results and build on them.

Thanks to the awesome @jessebett, one of the Neural ODEs authors, for helping us make this a reality, as well as @ChrisRackauckas, @oxinabox and @YingboMa for their utterly inimitable DE and ML expertise. Jesse and co. gave us an exciting taste of what’s possible in this emerging field, and we’re looking forward to seeing what new kinds of models you can build using these techniques.

Read all about it!


That’s wonderful, great work! I especially love how little code there is for tying everything together, which reflects a great design on the part of the packages involved. Now we just need fast reverse AD for every other package out there :slight_smile: