A Julia port of the Pytorch-Optim library. Purely pedagogical, but will maybe add Flux Interopability in the future. Stay tuned!
Pure Julia implementations of various Gradient Descent based Optimizers, along with visualizations on some functions listed here.
Visualization support is only for x where x isa AbstractVector && len(x)==2
Gradient Calculations based on ForwardDiff.jl
- Add basic tester functions
-
SGD - Construct optimization loop.
- Deploy visualizations using binder + Pluto.jl
- Other basic Optimizsrs
- One Advanced Optimiser
- Learning Rate and Weight decays
- Composability
- Research-y Optimisers