search

wujian16 / Cornell-MOE

A Python library for the state-of-the-art Bayesian optimization algorithms, with the core implemented in C++.
Other
261 stars 64 forks source link

number of restarts arguments in GradientDescentParameters #71

Open shalijiang opened 5 years ago

shalijiang commented 5 years ago

in struct GradientDescentParameters of moe/optimal_learning/cpp/gpp_optimizer_parameters.hpp, there are two parameter num_multistarts and max_num_restarts, how are they different?

108   //! number of initial guesses to try in multistarted gradient descent (suggest: a few hundred)
109   int num_multistarts;
110   //! maximum number of gradient descent iterations per restart (suggest: 200-1000)
111   int max_num_steps;
112   //! maximum number of gradient descent restarts, the we are allowed to call gradient descent.  Should be >= 2 as a minimum (suggest: 4-20)
113   int max_num_restarts;