An $O(\log OPT)$-approximation for covering and packing minor models of $θ_r$

Given two graphs $G$ and $H$, we define $\textsf{v-cover}_{H}(G)$ (resp. $\textsf{e-cover}_{H}(G)$) as the minimum number of vertices (resp. edges) whose removal from $G$ produces a graph without any minor isomorphic to ${H}$. Also $\textsf{v-pack}_{H}(G)$ (resp. $\textsf{v-pack}_{H}(G)$) is the maximum number of vertex- (resp. edge-) disjoint subgraphs of $G$ that contain a minor isomaorphic to $H$. We denote by $θ_r$ the graph with two vertices and $r$ parallel edges between them. When $H=θ_r$, the parameters $\textsf{v-cover}_{H}$, $\textsf{e-cover}_{H}$, $\textsf{v-pack}_{H}$, and $\textsf{v-pack}_{H}$ are NP-hard to compute (for sufficiently big values of $r$). Drawing upon combinatorial results in [Minors in graphs of large $θ_r$-girth, Chatzidimitriou et al., arXiv:1510.03041], we give an algorithmic proof that if $\textsf{v-pack}_{θ_r}(G)\leq k$, then $\textsf{v-cover}_{θ_r}(G) = O(k\log k)$, and similarly for $\textsf{v-pack}_{θ_r}$ and $\textsf{e-cover}_{θ_r}$. In other words, the class of graphs containing ${θ_r}$ as a minor has the vertex/edge Erdős-Pósa property, for every positive integer $r$. Using the algorithmic machinery of our proofs, we introduce a unified approach for the design of an $O(\log {\rm OPT})$-approximation algorithm for $\textsf{v-pack}_{θ_r}$, $\textsf{v-cover}_{θ_r}$, $\textsf{v-pack}_{θ_r}$, and $\textsf{e-cover}_{θ_r}$ that runs in $O(n\cdot \log(n)\cdot m)$ steps. Also, we derive several new Erdős-Pósa-type results from the techniques that we introduce.