The Complexity of Growing a Graph

Motivated by biological processes, we introduce here the model of growing graphs, a new model of highly dynamic networks. Such networks have as nodes entities that can self-replicate and thus can expand the size of the network. This gives rise to the problem of creating a target network $G$ starting from a single entity (node). To properly model this, we assume that every node $u$ can generate at most one node $v$ at every round (or time slot), and every generated node $v$ can activate edges with other nodes only at the time of its birth, provided that these nodes are up to a small distance $d$ away from $v$. We show that the most interesting case is when the distance is $d=2$. Edge deletions are allowed at any time slot. This creates a natural balance between how fast (time) and how efficiently (number of deleted edges) a target network can be generated. A central question here is, given a target network $G$ of $n$ nodes, can $G$ be constructed in the model of growing graphs in at most $k$ time slots and with at most $\ell$ excess edges (i.e., auxiliary edges $\notin E(G)$ that are activated and later deleted)? We consider here both centralized and distributed algorithms for such questions (and also their computational complexity). Our results include lower bounds based on properties of the target network and algorithms for general graph classes that try to balance speed and efficiency. We then show that the optimal number of time slots to construct an input target graph with zero-waste (i.e., no edge deletions allowed), is hard even to approximate within $n^{1-\varepsilon}$, for any $\varepsilon>0$, unless P=NP. On the contrary, the question of the feasibility of constructing a given target graph in $\log n$ time slots and zero-waste, can be answered in polynomial time. Finally, we initiate a discussion on possible extensions for this model for a distributed setting.