Network coding with unequal size overlapping generations

In this paper, we focus on content distribution applications with network coding in large networks where the network topology is unknown and no feedback is available. An unequal-size overlapping generation based network coding scheme is proposed. The sizes of generations are drawn from a degree distribution designed by and-or tree analysis. The overlapping of generations introduces redundancy and enables back substitution, which reduces the transmission overhead with little sacrifice of the computational complexity. Comparing with other overlapping based schemes (e.g. head-to-toe, random annex codes), our approach achieves a better overhead-complexity tradeoff.