Treedepth is a well-established width measure which has recently seen a resurgence of interest. Since graphs of bounded treedepth are more restricted than graphs of bounded tree- or pathwidth, we are interested in the algorithmic utility of this additional structure. On the negative side, we show with a novel approach that the space consumption of any (single-pass) dynamic programming algorithm on treedepth decompositions of depth d cannot be bounded by ${(2-ϵ)}^d·{log}^{O\left(1\right)}n$ for Vertex Cover, ${(3-ϵ)}^d·{log}^{O\left(1\right)}n$ for 3-Coloring and ${(3-ϵ)}^d·{log}^{O\left(1\right)}n$ for Dominating Set for any $ϵ>0$. This formalizes the common intuition that dynamic programming algorithms on graph decompositions necessarily consume a lot of space and complements known results of the time-complexity of problems restricted to low-treewidth classes. We then show that treedepth lends itself to the design of branching algorithms. Specifically, we design two novel algorithms for Dominating Set on graphs of treedepth d: A pure branching algorithm that runs in time $d^{O(d^2)}·n$ and uses space $O(d^{3}logd+dlogn)$ and a hybrid of branching and dynamic programming that achieves a running time of $O(3^{d}logd·n)$ while using $O(2^{d}dlogd+dlogn)$ space.

When can $t$ terminal pairs in an $m \times n$ grid be connected by $t$ vertex-disjoint paths that cover all vertices of the grid? We prove that this problem is NP-complete. Our hardness result can be compared to two previous NP-hardness proofs: Lynch's 1975 proof without the ``cover all vertices'' constraint, and Kotsuma and Takenaga's 2010 proof when the paths are restricted to have the fewest possible corners within their homotopy class. The latter restriction is a common form of the famous Nikoli puzzle \emph{Numberlink}; our problem is another common form of Numberlink, sometimes called \emph{Zig-Zag Numberlink} and popularized by the smartphone app \emph{Flow Free}.

The width measure treedepth, also known as vertex ranking, centered coloring and elimination tree height, is a well-established notion which has recently seen a resurgence of interest. Since graphs of bounded treedepth are more restricted than graphs of bounded tree- or pathwidth, we are interested in the algorithmic utility of this additional structure. On the negative side, we show that every dynamic programming algorithm on treedepth decompositions of depth~$t$ cannot solve Dominating Set with $O((3-\epsilon)^t \cdot \log n)$ space for any $\epsilon > 0$. This result implies the same space lower bound for dynamic programming algorithms on tree and path decompositions. We supplement this result by showing a space lower bound of $O((3-\epsilon)^t \cdot \log n)$ for 3-Coloring and $O((2-\epsilon)^t \cdot \log n)$ for Vertex Cover. This formalizes the common intuition that dynamic programming algorithms on graph decompositions necessarily consume a lot of space and complements known results of the time-complexity of problems restricted to low-treewidth classes. We then show that treedepth lends itself to the design of branching algorithms. This class of algorithms has in general distinct advantages over dynamic programming algorithms: a) They use less space than algorithms based on dynamic programming, b) they are easy to parallelize and c) they provide possible solutions before terminating. Specifically, we design for Dominating Set a pure branching algorithm that runs in time $t^{O(t^2)}\cdot n$ and uses space $O(t^3 \log t + t \log n)$ and a hybrid of branching and dynamic programming that achieves a running time of $O(3^t \log t \cdot n)$ while using $O(2^t t \log t + t \log n)$ space. Algorithms for 3-Coloring and Vertex Cover with space complexity $O(t \cdot \log n)$ and time complexity $O(3^t \cdot n)$ and $O(2^t\cdot n)$, respectively, are included for completeness.

This research establishes that many real-world networks exhibit bounded expansion, a strong notion of structural sparsity, and demonstrates that it can be leveraged to design efficient algorithms for network analysis. We analyze several common network models regarding their structural sparsity. We show that, with high probability, (1) graphs sampled with a prescribed s parse degree sequence; (2) perturbed bounded-degree graphs; (3) stochastic block models with small probabilities; result in graphs of bounded expansion. In contrast, we show that the Kleinberg and the Barabasi-Albert model have unbounded expansion. We support our findings with empirical measurements on a corpus of real-world networks.

We prove that for every positive integer $r$ and for every graph class $\mathcal G$ of bounded expansion, the $r$-Dominating Set problem admits a linear kernel on graphs from $\mathcal G$. Moreover, when $\mathcal G$ is only assumed to be nowhere dense, then we give an almost linear kernel on $\mathcal G$ for the classic Dominating Set problem, i.e., for the case $r=1$. These results generalize a line of previous research on finding linear kernels for Dominating Set and $r$-Dominating Set. However, the approach taken in this work, which is based on the theory of sparse graphs, is radically different and conceptually much simpler than the previous approaches. We complement our findings by showing that for the closely related Connected Dominating Set problem, the existence of such kernelization algorithms is unlikely, even though the problem is known to admit a linear kernel on $H$-topological-minor-free graphs. Also, we prove that for any somewhere dense class $\mathcal G$, there is some $r$ for which $r$-Dominating Set is W[$2$]-hard on $\mathcal G$. Thus, our results fall short of proving a sharp dichotomy for the parameterized complexity of $r$-Dominating Set on subgraph-monotone graph classes: we conjecture that the border of tractability lies exactly between nowhere dense and somewhere dense graph classes.