Release v1.2.57

PLS overhaul
Priori this version, the PLS algorithm was orginally made for multi-variate output, thus requiring a SVD at each iteration (aka component computation). But in the abcrf setting, we only predict one parameter at a time, thus univariate and much lighter version of the PLS algorithm (and faster by several order of magnitude : just ONE inversion required).

Release v1.2.56

Fix #83

Release v1.2.54

Typo

Release v1.2.53

Minor : see #81

Release v1.2.52

Minor : Fix missing multi-target output for local weights in parameter estimation.

Release v1.2.50

Multi target support
In standalone this means you can read several target (one for each line) in statobs file. Works for model choice and parameter estimation.
Fix also #79

Release v1.2.49

Fix cosmetic in lda output.

Release v1.2.48

Fixed major bug when pop number is greater or equal than 10

Release v1.2.47

• Fixed bug in -g (group) parsing for standalone

Release v1.2.35

Fix a very rare crash in heavy multi-threaded environment.