gaussian_learning_rate_function_utilities.f90 Source File

!! author: Oscar Garcia-Cabrejo
!! date: 03/18/2025
!! version: 0.1
!! This module defines a class that represents the gaussian learning rate function
module gaussian_learning_rate_function_utilities
!! This module defines a class that represents the gaussian learning rate function
    use precision_utilities, only: wp;
    use learning_rate_function_base_utilities, only: learning_rate_function_base;
    implicit none;
    private;
    !
    type,extends(learning_rate_function_base) :: gaussian_learning_rate_function
        private
            real(kind=wp) :: lambda0,tau
        contains
            procedure,public :: set_parameters
            procedure,public :: calculate => calculate_gaussian_function
    end type gaussian_learning_rate_function
    !
    public :: gaussian_learning_rate_function
    !
    contains
        subroutine set_parameters(learning_rate_fn,lambda0,tau)
            class(gaussian_learning_rate_function) :: learning_rate_fn
            real(kind=wp) :: lambda0,tau
            !
            learning_rate_fn%lambda0=lambda0;
            learning_rate_fn%tau=tau;
    
        end subroutine set_parameters
        !
        function calculate_gaussian_function(learning_rate_fn,iteration) result(alpha)
            class(gaussian_learning_rate_function) :: learning_rate_fn
            !! An `learning_rate_function_base` object
            real(kind=wp),intent(inout) :: iteration
            !! A real variable with the current iteration
            real(kind=wp) :: alpha
            !! A real variable with the learning rate
            real(kind=wp) :: time_factor
            !
            time_factor=-(iteration**2/(2.0_wp*learning_rate_fn%tau**2));
            alpha=max(learning_rate_fn%lambda0*dexp(time_factor),0.01_wp);
    
        end function calculate_gaussian_function
!
end module gaussian_learning_rate_function_utilities