Pareto Optimal Schemes in Coded Caching: Uncoded Prefetching

The problem of coded caching was introduced by Maddah-Ali and Niesen and has been extensively studied in recent years. The problem is fundamentally a multi-objective optimization problem where the rates achieved for each demand type is of interest and Pareto optimality is a natural framework. Under the constraint that the placement phase is uncoded, Yu et al. introduced the YMA scheme which was shown to be universal for all demand types. Vijith et al. showed that there are no universal schemes when coded placement is permitted and introduced the problem of finding Pareto optimal schemes. In this paper we study the possibility of finding schemes that dominate the YMA scheme and demonstrate, rather surprisingly, that they continue to operate at the Pareto optimal frontier of coded caching for (N, K) cache networks when $K$ ≤ 3. We introduce new lower bounds which partially characterize the tradeoffs between different demand types.