An Experimental Study of Recommendation for Wireless Edge Caching

Wei, Dasen; Han, Shengqian

doi:10.1109/iccc55456.2022.9880741

Cited by 4 publications

(1 citation statement)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [23][24][25], deep reinforcement learning was used to simultaneously optimize caching and recommendation systems. In [26][27][28][29], the focus was not only on increasing the hit ratio, but also on improving the quality of experience to increase user satisfaction. In [30], instead of recommending a single item, a set of items is recommended.…”

Section: Introductionmentioning

confidence: 99%

Content Caching Strategies With On-Device Recommendation Systems in Wireless Caching Systems

Rim

2024

IEEE Access

View full text Add to dashboard Cite

Wireless network traffic is exploding, mainly due to the growth of video streaming services. To cope with the increase of mobile traffic in wireless networks, techniques for installing content caches in base stations or devices are being investigated. Although the amount of total content is huge, the capacity of a cache installed in base stations or devices is limited, so efficient caching methods are needed to improve the hit ratio. To this end, we can consider content recommendation systems on devices. Since many users tend to select and watch videos from recommended content, a cache can improve its hit ratio by storing content that is more likely to be recommended on each device. This paper discusses how caching systems differ when devices recommend content versus when they do not. It also discusses how caching systems differ when recommendations are made based on cached content versus when they are not. Content with high average preferences should be cached without recommendations, while the caching system should take personal preferences into account when recommending content personally preferred by each device. This is especially true for cache-independent recommendation systems, since each device will recommend personal favorites regardless of cached content. If the cache size is very large compared to the number of recommended contents, the consideration of cached content in recommendation systems may become less important, since much personal content should be cached anyway. The simulation results show that caching schemes with recommendation systems can differ significantly from those without.

show abstract