Core Concepts
By leveraging side-information and reconfigurable antennas at receivers, blind interference alignment schemes can achieve higher degrees of freedom in wireless MapReduce networks compared to conventional interference management approaches.
Abstract
The paper explores how blind interference alignment (BIA) schemes can take advantage of side-information in wireless MapReduce applications. In the considered setting, receivers have reconfigurable antennas and channel knowledge, while transmitters lack channel state information.
The key contributions are:
Establishing a connection between the MapReduce problem and a corresponding vector broadcast channel with groupcast messages (BCGM) setting. This allows the authors to first solve the DoF of the BCGM setting, which is a more tractable problem.
Developing a new BIA scheme for the BCGM setting that requires not only intra-message alignment (as in conventional BIA), but also inter-message alignment to exploit the side-information structure. This new coding scheme achieves the optimal sum-DoF for the BCGM setting.
Mapping the BCGM solution back to the original MapReduce setting to obtain the DoF characterization. The authors introduce an intermediate "Unicast with Side Information" (USI) setting to facilitate this mapping.
The results show that by leveraging side-information and reconfigurable antennas, the proposed BIA schemes can achieve higher DoF compared to prior works that either assumed perfect CSIT or did not consider side-information.