Stan Math Library  2.10.0
reverse mode automatic differentiation
grad.hpp
Go to the documentation of this file.
1 #ifndef STAN_MATH_REV_CORE_GRAD_HPP
2 #define STAN_MATH_REV_CORE_GRAD_HPP
3 
9 #include <vector>
10 
11 namespace stan {
12  namespace math {
13 
30  static void grad(vari* vi) {
31  // simple reference implementation (intended as doc):
32  // vi->init_dependent();
33  // size_t end = var_stack_.size();
34  // size_t begin = empty_nested() ? 0 : end - nested_size();
35  // for (size_t i = end; --i > begin; )
36  // var_stack_[i]->chain();
37 
38  typedef std::vector<vari*>::reverse_iterator it_t;
39  vi->init_dependent();
40  it_t begin = ChainableStack::var_stack_.rbegin();
41  it_t end = empty_nested()
42  ? ChainableStack::var_stack_.rend() : begin + nested_size();
43  for (it_t it = begin; it < end; ++it) {
44  (*it)->chain();
45  }
46  }
47 
48 
49  }
50 }
51 
52 #endif
static bool empty_nested()
Return true if there is no nested autodiff being executed.
The variable implementation base class.
Definition: vari.hpp:30
static void grad(vari *vi)
Compute the gradient for all variables starting from the specified root variable implementation.
Definition: grad.hpp:30
static size_t nested_size()
Definition: nested_size.hpp:10
void init_dependent()
Initialize the adjoint for this (dependent) variable to 1.
Definition: vari.hpp:98
static std::vector< ChainableT * > var_stack_

     [ Stan Home Page ] © 2011–2016, Stan Development Team.