dl::optim::GradientDescent Class Reference

Inheritance diagram for dl::optim::GradientDescent:

Collaboration diagram for dl::optim::GradientDescent:

Public Member Functions
	GradientDescent (std::map< std::string, dl::TensorRef > &parameters, float learnrate=0.001f)
virtual void	step (dl::TensorPtr &loss) override

Detailed Description

Definition at line 11 of file gradientdescent.hpp.

Constructor & Destructor Documentation

GradientDescent()

dl::optim::GradientDescent::GradientDescent ( std::map< std::string, dl::TensorRef > &  parameters, float  learnrate = 0.001f  )

inlineexplicit

Definition at line 17 of file gradientdescent.hpp.

                : dl::Optimizer(), parameters(parameters), learnrate(learnrate) {}

Member Function Documentation

step()

virtual void dl::optim::GradientDescent::step ( dl::TensorPtr &  loss )

inlineoverridevirtual

Implements dl::Optimizer.

Definition at line 20 of file gradientdescent.hpp.

                                                      {
            loss->backward();
            for (auto&& [_, tensor] : parameters) {
                auto& gradient = tensor.get()->gradient();
                assert(gradient != nullptr);
                tensor.get() = tensor.get()->add(gradient->mul(dl::constant(-learnrate, gradient->device())));
            }
        }

---

The documentation for this class was generated from the following file:

• dl/learning/optimizers/gradientdescent.hpp