Qifa Yan

Online Coded Caching With Random Access

In this letter, we consider a caching network with two characteristics: first, the operations of users’ caches have to be performed on the fly, and second, each user has access to the system with a known probability. Particularly, we intend to characterize approximately the optimal long-term average rate of the shared link. By utilizing the knowledge of eviction in the server and applying the technique of coded caching to the active users in each time, we propose and analyze a new scheme called coded outdated file eviction for random access (coded OER). Furthermore, we present a tight upper bound on the performance of the proposed coded OER scheme and prove that it is approximately optimal.

On the greedy coded caching scheme

Coded caching is a technique that receives much attention in the network literature, since it is able to achieve significant gains compared to the conventional uncoded caching schemes. In this paper, we first identify a family of placement and delivery schemes. Then we derive a lower bound on the average rate for the identified schemes. Based on the clue in deriving the lower bound, a new delivery algorithm is proposed to pursue full multicast adaptively, which can exploit more cooperative opportunities compared to the known schemes. Particularly, the proposed scheme has the advantage that, when the file size is not large enough, the proposed scheme is still able to achieve a reasonable coding gain. We perform simulations to validate our design.

An Improved Algorithm for Testing the Membership of Gaussian Multiple-Access Channel Capacity Region

In this letter, we investigate the problem of testing the membership of Gaussian multiple-access channel capacity region (GMACR), which involves 2N - 1 inequalities, where N is the number of senders. Given a fixed rate tuple and positive power constraint, we propose a simple but efficient algorithm to test the membership of the GMACR. It is unveiled that the algorithm can be accomplished in O(N log2 N) operations, which is much more efficient than the approach by Zhang et al., whose complexity bound is larger than O(N5).