Lower bounds for kernelization

Kernelization is the process of transforming the input of a combinatorial decision problem to an equivalent instance, with a guarantee on the size of the resulting instances as a function of a parameter. Recent techniques from the field of fixed parameter complexity and tractability allow to give lower bounds for such kernels. In particular, it is discussed how one can show for parameterized problems that these do not have polynomial kernels, under the assumption that coNP⊆ NP/poly.