Merge branch 'master' of github.com:mz2/peer

TODO

@@ -1,9 +1,12 @@
-1) Test that the covariates functionality works propperly (Leo, there are some comments in the codes also)
-2) Deal with all the debugging output. Where shall these messages go in the future?
-3) R wrapper, I just don't understand how R works, the lib is build; found this here (http://www.swig.org/Doc1.3/R.html#R_nn6) which does not help me.. 
-4) Priors (pa,pb) have setters but no getters, probably not that critical though..
-5) Documentation, doxygen? 
+1) Test that the covariates functionality works propperly (Leo, there are some comments in the codes also) (Leo)
+2) Deal with all the debugging output. Where shall these messages go in the future? Verbosity. (Oli)
+3) R wrapper, I just don't understand how R works, the lib is build; found this here (http://www.swig.org/Doc1.3/R.html#R_nn6) which does not help me.. (Matias)
+4) Priors (pa,pb) have setters but no getters, probably not that critical though.. (Oli)
+5) Documentation, doxygen? (Matias)
 6) Propper build system using autoconfigu or similar.
    Matias: we need to get the python support in also. Either we use the python distutils for this (I currenlty do) or we need python-configure to figure out where python is  etc.
    With autoconfig that is all reasonably easy. Maybe "shogun" (http://www.shogun-toolbox.org/) has some things to steal from.
-   I copied the typempas for R and python from them also, foud this very useful in many ways.
+   I copied the typempas for R and python from them also, foud this very useful in many ways. (Matias)
+7) Check for memory leaks
+8) Stress tests for large matrices - memory usage, speed for CPP/Py/R versions (Leo)
+9) Convergence - from lower bound, not number of iterations (Oli)