erfc: Complementary Gaussian Error Function

In vcerqueira/tsensembler: Dynamic Ensembles for Time Series Forecasting

Description

Erfc stands for the Complementary Gaussian Error Function. This mathematical formula can be used as a squashing function. Consider x a numeric vector representing the squared error of base models in a given observation. By applying the erfc function on the error, the weight of a given model decays exponentially as its loss increases.

Usage

erfc(x, alpha = 2)

Arguments

x	A numeric vector. The default value for the parameter lambda presumes that x is in a 0–1 range. In the scope of this package, this is achieved using normalize function;
alpha	parameter used to control the flatness of the erfc curve. Defaults to 2.

Value

The complementary Gaussian error

References

Cerqueira, Vitor; Torgo, Luis; Oliveira, Mariana, and Bernhard Pfahringer. "Dynamic and Heterogeneous Ensembles for Time Series Forecasting." Data Science and Advanced Analytics (DSAA), 2017 IEEE International Conference on. IEEE, 2017.

Examples

## Not run: 
  erfc(.1)
  erfc(c(.1, .7))

## End(Not run)

vcerqueira/tsensembler documentation built on Oct. 28, 2020, 11:46 p.m.