########## Version 0.2.0 (September 2026) ###############

The following changes were made in Version 0.2.0 of R package ProbSVMs:
 
1) The default value of the hyper-parameter eta in the predictPVM function
   changed from 15 to the string adjtogrid. With this default the value of
   eta is automatically computed from the grid specification in such way that
   it is large enough to ensure that the undesirable deviations dominate the 
   desirable ones in the objective to be minimized. This cond ition is necessary
   to inforce the consistency properties of the resulting PVM estimators.

2) The default value of the argument withbias in in the trainPVM and trainSVM
   functions changed from length(levels(y))==2 to FALSE. This way in all problems
   the default is consistently not include bias terms, instead of using bias
   terms by defaul to 2-class problems and dropping them for problems with more
   than two groups, as it was the case in version 0.1.0.  

3) A bug in the normalization (standardization) routine of functions predictPVM 
   and predictSVM was corrected. In version 0.1.0 when different data was used for
   prediction/estimation than the data used for training, the normalization 
   (standardization) used in the newdata did not always ensured that the same
   scaling constants was used for the training and the newdata. This problem
   was corrected in version 0.2.0 were the same constants used for normalizing
   the training data ara always used for normalizing any newdata used for prediction
   or estimation.        

4) In the function predictPVM for problems with two classes the LP based boubded-variables
   simplex algorithm was replaced by an equivalent, but faster, optimization algorithm
   specialized for two-class problems.    
