Package: evreg
Title: Evidential Regression


Changes in version 1.1.1

* The changes made in version 1.1.0 concerning the loss function of ENNreg and its gradient  could result in performance degradation. These changes have been reverted, so that results reported in the IEEE TFS paper can be reproduced.


Changes in version 1.1.0

* An error in the formula used in function "Bel" has been corrected.
* The calculation of the loss function and its gradient in ENNreg has been corrected.
* Function "intervals" now returns the average plausibility of the belief intervals.


Changes in version 1.0.4

* A bug in function "ENNreg_init" has been corrected.
* Function ENNreg now returns the list of all hyperparameters.
* References have been updated.


Changes in version 1.0.3

* The line "@aliases evreg-package" has been added to the file "evreg.R" to facilitate access to the documentation.


Changes in version 1.0.2

* Function "combination_GRFN" now returns a list with the combined GRFN and the degree of conflict.
* Function "combination_GRFN" has been modified to produce correct results when GRFN1$sig=0 or  GRFN2$sig0, and when GRFN1$h=GRFN2$h=0.
* Function pl_contour has been modified to produce correct results when the GRFN is vacuous and one of the bounds in infinite.


Changes in version 1.0.1

* Correction of a bug in function "predict.ENNreg" that caused an error when the test set had only one instance.
* The cross-validation subsets are sampled differently in function "ENNreg.cv" to avoid empty subsets when the initial sample is small.
* Correction of typos in the documentation.