Planned changes in next versions:
----------------------------------------
-verify if plot and importance functions still work correctly after upgrade to 1.7.5 (already verified for summary function, but still need to check LaTeX parameters in this function)
-adapt combination methods to handle varying number of sub-ensembles (specified in 'algorithms' parameter). rbga and soma are already done.
-add parallel parameter to hybridEnsemble(). CVhybridEnsemble has a parallel parameter already.
-add other performance measures (now only AUC) in diversity calculation

Change log:
----------------------------------------
* Version 1.7.9: March 8, 2023
-Add ... to importance function

* Version 1.7.8: May 9, 2022
-Import new pacakges for CRAN 

* Version 1.7.8: February 15, 2017
-enabled soma

* Version 1.7.7: December 16, 2016
-turned off refitting model on training + validation data after validating to obtain better correspondence between weights and models
-fixed reporting of diversity for predSB

* Version 1.7.6: September 5, 2016
-bug fix for KNN. Renamed everything to KN.
-added more diversity output

* Version 1.7.5: September 4, 2016
-changed name of diversity parameter to 'diversity' in  CVhybridEnsemble() and added accuracy in the output
-comment out all combination methods except rbga
-added parallel parameter to CVhybridEnsemble()
-adapted defaults of rbga to handle varying number of sub-ensembles
-added a parameter, called algorithms, in hybridEnsemble() and CVhybridEnsemble() to be able to select a subset of algorithms. A minimum of two algorithms is required.
-added ninth base classifier: Bagged Naive Bayes to hybridEnsemble(), and predict()
-added skip parameter (nnet) to hybridEnsemble() and CVhybridEnsemble() with tuning option
-nearest neighbors (in hybridEnsemble() and predict()): added scaling to [0,1]
-added error checking for dependent variable (can only be 0 or 1) in hybridEnsemble()
-improved verbose printing
-resolved inconsistencies in documentation

* Version 1.2.0: December, 15, 2015
-fixed a bug related to the filter parameter
-fixed a minor bug in passing on the optimal k for kNN 
-CVhybridEnsemble: compute diversity (1 minus absolute value of mean of pairwise correlations) for all sub-ensembles and meta ensemble, along with AUC 
-adapts summary and plot functions to handle predict.all
-parameter predict.all in CVhybridEnsemble function
-parameter calibrate to use percentile ranks instead of calibration
-parameter predict.all in predict function to yield prediction vectors of sub-ensembles and base classifiers
-fixed a bug in the bootstrapping procedure for KNN
-fixed a bug in the improvement of the bootstrapping for LR, NN, SV, introduced in version 1.0.0

* Version 1.0.0: May, 26 2015
-added a seventh base classifier: rotation forest
-added an eight base classifier: bagged nearest neighbors
-added parameters for size of sub-ensembles
-added oversampling to alleviate problems related to class imbalance and subsequent subsetting
-added a filter parameter to remove near constants that often produced problems in subsetting.
-added error handling
-added formal automated tests
-refactored for faster and shorter code
-changed to [-1,1] scaling in neural networks
-made sure that no (near) constants can be produced in bootstrap procedures
-corrected typesetting issues in documentation

* Version 0.1.1: March 2014
- Fixed small bug that occurred in rare cases

* Version 0.1.0: December 2013
- Package submitted to CRAN with main functions hybridEnsemble, predict, importance, CVhybridEnsemble, plot, summary