Preprint out for the paper: DEAP Cache: Deep Eviction Admission and Prefetching for Cache!
In this paper, we propose a DL based approach to tackle the problem of Cache Replacement. This is the first time an approach has tried learning all the three policies: Admission, Prefetching and Eviction. Unlike, previous methods which relied on past statistics for carrying out cache replacement, we predict future statistics (frequency and recency) and then use an online RL-algorithm for eviction.
Checkout the preprint here.