Online Spatiotemporal Popularity Learning via Variational Bayes for Cooperative Caching

Abstract

Herein, we focus on an end-to-end design of a proactive cooperative caching strategy for a multi-cell network. The design is challenging as it involves two interrelated problems: the ability to predict future content popularity and to meet network operation characteristics. To this end, we first formulate a cooperative content caching in order to optimize the aggregated network cost for delivering contents to users. An efficient proactive caching policy requires an accurate prediction of timevarying content popularity. Content popularity has temporal and spatial dependencies and therefore, we develop a probabilistic dynamical model for content popularity prediction by exploiting its spatiotemporal correlations. To achieve an accurate tracking and prediction of content popularity evolution, the proposed dynamical model is non-linear and incorporates non-Gaussian distributions. We use Variational Bayes (VB) approach for estimating the model parameters. The VB provides mathematical tractability. We then develop an online VB method that works with streaming data where content request arrives sequentially. Using extensive simulations study on a real-world dataset, we show that our online VB based dynamical model provides improved performance compared to conventional content caching policies.

Trend-Aware Proactive Caching via Tensor Train Decomposition: A Bayesian Viewpoint

“…Third, the time factor, Z t , is a full matrix which enhances the expressiveness power of the model. We also note that the CP tensor factorization method proposed in [27] for Poisson data is a special case of the formulation in (1), where the time factor matrix is diagonal. A graphical view of decomposition model (1) is shown in Fig.…”

“…In our previous work [26], we proposed a Bayesian model based on matrix factorization approach to capture the interactions among the contents in a time varying scenario. More recently in [27], we introduced a dynamical model using a CAN-DECOM/PARAFAC (CP) tensor decomposition to capture content-location interactions.…”

“…Parallelly, there has been a great interest in exploring optimal proactive cooperative caching policy and performance analysis for an efficient use of network resources. The vast majority of work, however, focused on static scenarios or designed policies with fixed resource dimensioning [13], [28], [29], and [27]. A more efficient and realistic scheme is to design a policy that adaptively adjusts network resources according to the dynamic of content popularity.…”

“…Moreover, a simple-to-implement approximation method based on the Kalman smoother and VB is derived to infer the parameters of Bayesian dynamical model. The Bayesian trend-aware train decomposition model developed in this work has a more expressiveness power than the non-trend CP decomposition model in [27]. Therefore, the VB algorithm designed here is more sophisticated than the one in [27].…”

“…The Bayesian trend-aware train decomposition model developed in this work has a more expressiveness power than the non-trend CP decomposition model in [27]. Therefore, the VB algorithm designed here is more sophisticated than the one in [27]. • To provide an end-to-end caching system design, a joint routing and content placement policy is designed for cooperative caching to minimize network cost by adaptively utilizing the network resources, i.e., cache memory and link bandwidth, in time-varying scenario.…”