FastAD is a C++ implementation of automatic differentiation both forward and reverse mode.
bind_cache
instead of bind
to get rid of all the if constexpr
: saves compile-timeBiggest change is the new support for vector and matrix-like shapes. Reverse-mode now supports vectorized operations and uses Eigen as a dependency to carry out all matrix computations. Removed jacobian and hessian for the time being since they don't work so well with the new API.
Release date: 2019-12-23.
This is the first release of FastAD. See README for all features and usage examples.