Interference is a key property of wireless communications due to the broadcasting nature of wireless links. The design of wireless networks needs to put interference management into consideration. Traditionally, interference management is done by partitioning the whole network into orthogonal non-interfering channels via time- or frequency-division multiplexing. While orthogonalization significantly reduces the complexity of the design and implementation of wireless networks, it also introduces artificial restriction and leads to suboptimal performance. This thesis is devoted to the design and analysis of interference management from a cross-layer perspective. The key to increase spectrum efficiency of a wireless network is to treat the entire network as a channel rather than viewing them as a set of separate links. Based on this idea, we propose three interference management schemes and evaluate the fundamental limits associated with them. We use the notions of both conventional and generalized degrees of freedom (DOF), which are two widely-used approximations of channel capacity, as merits to evaluate and compare the performance improvement brought by the interference management schemes. The thesis consists of four main results. First, we consider a multiple-input-multiple-output (MIMO) 2-suer cognitive radio system in an information theoretic setting where some messages are made available, by a genie, to some nodes (other than the intended nodes) non-causally, noiselessly, and for free. We find the DOF region of this system and show that this region is larger than the one without cognitive message sharing. Our results also show that in general it may be more beneficial, in terms of sum DOF, for a user to have a cognitive transmitter than to have cognitive receiver. Second, we consider a MIMO Gaussian interference channel with user cooperation, including cooperation at transmitters only, at receivers only, and at transmitters as well as receivers. We find the DOF region of this system and obtain a negative result that allowing users to cooperate does not enlarge the DOF region of this channel. Third, we explore the capacity and generalized degrees of freedom (GDOF) of a 2-user Gaussian X channel, i.e. a generalization of the 2-user interference channel where there is an independent message from each transmitter to each receiver. We provide the GDOF characterization of the channel under a symmetric setting. We also identify the regime where interference alignment is helpful so that the X channel has a higher capacity than the underlying symmetric interference channel. We further extend the noisy interference capacity characterization previously obtained for the interference channel to the X channel. Lastly, we study the effect of the absence of channel knowledge for MIMO networks. In particular, we assume perfect channel state information at the receivers and no channel state information at the transmitter(s). We provide the characterization of the DOF region for a 2-user MIMO broadcast channel. We then use the result of the broadcast channel to find the DOF region for some special cases of a 2-user MIMO interference channel.