Changes in 1.2.1:

	* add a new function 'hlaResource'
	* minor improvement


Changes in 1.2.0:

	* optimize the C code: replace some intensive 'exp' and 'log' calculations by precalculated values, roughly 2.5x ~ 4x speedup
	* optimize the C code of the hamming distance calculation between genotypes and two haplotypes by SSE2 intrinsics, ~50% faster for large datasets
	* the code optimization guarantees the same training and predictive accuracies as version 1.1.0
	* add a new function 'hlaPredMerge' to merge predictions from multiple models via voting
	* add assembly information (like hg19) to the objects of 'hlaAttrBagClass' and 'hlaAttrBagObj'
	* remove unused row names in the function 'hlaCompareAllele'
	* add an argument 'stop.cluster' to 'hlaParallelAttrBagging'
	* allow imputing HLA alleles in parallel at a single locus
	* allow matching SNPs by positions only
	* improve the function 'hlaPublish'
	* improve documents


Changes in 1.1.0:

	* add genome assembly information (such like hg18, hg19)
	* allow class majority voting in 'predict.hlaAttrBagClass'
	* add an autosave function to 'hlaParallelAttrBagging'
	* add a new function 'hlaPublish'
	* allow additional information in the objects of 'hlaAttrBagClass' and 'hlaAttrBagObj'
	* improvement in mismatching alleles for allelic strands
	* update citation information
	* change 'hlaModelfromObj' to 'hlaModelFromObj'
	* document improvements


Changes in 1.0.0:

	* use 'useDynLib()' in NAMESPACE instead of 'library.dynam()'
	* increase the limit of the possible number of SNPs in a single classifier up to 256
	* update citation information
	* add a vignette


Changes in 0.9.4:

	* document improvements
	* bug fix: passing only one individual to 'predict.hlaAttrBagClass'


Changes in 0.9.3:

	* add a load balancing function to 'hlaParallelAttrBagging'
	* add 'hlaUniqueAllele'
	* add an argument 'match.pos' to the functions 'predict.hlaAttrBagClass', 'hlaGenoCombine', 'hlaGenoSwitchStrand'
	* add an argument 'with.pos' to the function 'hlaSNPID'
	* give more warnings about missing SNPs in 'predict.hlaAttrBagClass'
	* avoid segmentation fault in 'predict.hlaAttrBagClass' when all SNPs are missing.
	* a warning is given for strand mismatching when calling 'hlaGenoSwitchStrand'
	* document improvements


Changes in 0.9.2:

	* minor fixes


Version 0.9.1:

	* first public release of HIBAG
