For those interested, the JuliaSim product demo which ran just after this talk in the Julia Computing Sponsor Talk can be found at th-cam.com/video/teqS3O362Ts/w-d-xo.html . Still pre-beta, but I hope to be sharing details about a launch product soon. Thanks for the feedback community!
- Vision: scalable cloud backend (distributed, gpu, parallel workloads) + transfer learning + GUIs... for each domain. - Just use Julia instead of optimizing MATLAB/R/Python workflows in C. - Speed comes from Julia, but moreover deep scientific computing and modeling ecosystem. - Embed neural networks as the missing params in differential equations. - Neural nets alone fail on stiff equations, so a hybrid approach is needed. - Using pretrained surrogate models results in ~ up to 700x speedup.
For those interested, the JuliaSim product demo which ran just after this talk in the Julia Computing Sponsor Talk can be found at th-cam.com/video/teqS3O362Ts/w-d-xo.html . Still pre-beta, but I hope to be sharing details about a launch product soon. Thanks for the feedback community!
- Vision: scalable cloud backend (distributed, gpu, parallel workloads) + transfer learning + GUIs... for each domain.
- Just use Julia instead of optimizing MATLAB/R/Python workflows in C.
- Speed comes from Julia, but moreover deep scientific computing and modeling ecosystem.
- Embed neural networks as the missing params in differential equations.
- Neural nets alone fail on stiff equations, so a hybrid approach is needed.
- Using pretrained surrogate models results in ~ up to 700x speedup.
My favorite talk from this year's JuliaCon :)
Very interesting talk and right on the money