Smoothing.cpp File Reference

#include "Smoothing.h"
#include "autopas/tuning/searchSpace/Evidence.h"
#include "autopas/utils/Math.h"

Include dependency graph for Smoothing.cpp:

Functions
std::tuple< std::vector< double >, bool >	calculateWeightsSimple (const std::vector< autopas::Evidence > &points, size_t pointsPerEstimation, size_t maxDistFromIntervalEdge)
	Calculates the weights for the k-neighbors of the last point in the vector using the tri-cube function.
double	calculateYFitSimple (const std::vector< autopas::Evidence > &points, size_t pointsPerEstimation, const std::vector< double > &weights)
	Calculates the smoothed y-value of the last point in the vector.

Detailed Description

Author

F. Gratl

Date

23/11/2020

Function Documentation

calculateWeightsSimple()

std::tuple< std::vector< double >, bool > calculateWeightsSimple	(	const std::vector< autopas::Evidence > &	points,
		size_t	pointsPerEstimation,
		size_t	maxDistFromIntervalEdge
	)

Calculates the weights for the k-neighbors of the last point in the vector using the tri-cube function.

If the sum of weights is <= 0 all neighbors are at the same x coordinate as the point and smoothing will not change the point. This case is indicated by the bool in the return tuple.

Parameters

points	Sorted collection of observations.
pointsPerEstimation	Number of neighbors to consider for smoothing.
maxDistFromIntervalEdge	Largest distance between the point that shall be fitted and its neighbors.

Returns

Tuple of a vector containing the weights for the neighbors and a bool indicating whether the fitting should be used.

calculateYFitSimple()

double calculateYFitSimple	(	const std::vector< autopas::Evidence > &	points,
		size_t	pointsPerEstimation,
		const std::vector< double > &	weights
	)

Calculates the smoothed y-value of the last point in the vector.

The fitted value is the sum of projections of the y-values of the neighbors in the chosen interval. Each projection is the sum of the respective weight and the proportion of the residuals of the weighted sum of squared residuals.

Parameters

points
pointsPerEstimation
weights

Returns