The early processing of olfactory information in the olfactory bulb of rodents is highly dynamic. On long time scales the connectivity of its neuronal network evolves persistently - even in adult animals - through synaptic rewiring and the addition and removal of neurons. On short time scales the neuronal activity exhibits pronounced population rhythms reflecting the coherent activity of large ensembles of neurons. Starting from various experimental observations, we have developed simple computational models of this network evolution that suggest that cortical feed-back controls the adaptive formation of subnetworks through which cortex then modifies sensory processing by the olfactory bulb. The gamma-rhythms of different such subnetworks can have different frequencies. We show that even uncorrelated noise can synchronize these population rhythms.