Estimating divergence functionals and the likelihood ratio by penalized convex risk minimization. X. Nguyen, M. Wainwright and M.I. Jordan, NIPS 2007.
In Section 6 (Simulation) M1 was used to refer to the case
where G is the RKHS, and M2 the case where \log G is the RKHS.
It should be the reverse: M1 should denote the case
where \log G is the RKHS; M2 the case where G is the RKHS.