Follow the gradient based on the partial derivatives. Multiply by the current step size. If the score gets worse, reduce the step size. If the score gets better, increase the step size if we are sufficiently far from a score of zero. If the score reaches the threshold, quit.
Public Member Functions | |
virtual std::string | get_type_name () const |
virtual ::IMP::VersionInfo | get_version_info () const |
virtual Float | optimize (unsigned int max_steps) |
void | set_step_size (Float t) |
Set the step size. | |
void | set_threshold (Float t) |
Set the minimum gradient threshold. | |
SteepestDescent (Model *m=NULL) | |
Friends | |
template<class T > | |
void | IMP::internal::unref (T *) |