Low-complexity Rank-Efficient Tensor Completion For Prediction And Online Wireless Edge Caching

TL;DR

This paper introduces a low-complexity, rank-efficient tensor completion method for improving online wireless edge caching, reducing computational overhead while enhancing prediction accuracy and cache hit rates.

Contribution

It proposes a novel multi-rank update tensor completion algorithm tailored for large-scale, dynamic edge caching scenarios, enabling real-time predictions with lower computational costs.

Findings

Lower reconstruction errors compared to recent FW algorithms

Reduced computational overhead in tensor completion

Improved cache hit rates with the proposed method

Abstract

Wireless edge caching is a popular strategy to avoid backhaul congestion in the next generation networks, where the content is cached in advance at base stations to serve redundant requests during peak congestion periods. In the edge caching data, the missing observations are inevitable due to dynamic selective popularity. Among the completion methods, the tensor-based models have been shown to be the most advantageous for missing data imputation. Also, since the observations are correlated across time, files, and base stations, in this paper, we formulate the cooperative caching with recommendations as a fourth-order tensor completion and prediction problem. Since the content library can be large leading to a large dimension tensor, we modify the latent norm-based Frank-Wolfe (FW) algorithm with towards a much lower time complexity using multi-rank updates, rather than rank-1 updates in literature. This significantly lower time computational overhead leads in developing an online caching algorithm. With MovieLens dataset, simulations show lower reconstruction errors for the proposed algorithm as compared to that of the recent FW algorithm, albeit with lower computation overhead. It is also demonstrated that the completed tensor improves normalized cache hit rates for linear prediction schemes.