Automatic Differentiation
 
Loading...
Searching...
No Matches

◆ WolfeInfo() [1/3]

template<typename ObjFun , typename Theta0 , typename AInit , typename ThetaGradF >
stan::math::internal::WolfeInfo::WolfeInfo ( ObjFun &&  obj_fun,
AInit &&  a_init,
Theta0 &&  theta0,
ThetaGradF &&  theta_grad_f 
)
inline

Construct WolfeInfo with a consistent (a_init, theta_init) pair.

When the caller supplies a non-zero theta_init, the corresponding a_init = Sigma^{-1} * theta_init must be provided so that the invariant theta = Sigma * a holds at initialization. This avoids an inflated initial objective (the prior term -0.5 * a'*theta would otherwise vanish when a is zero but theta is not).

Definition at line 520 of file wolfe_line_search.hpp.