Given a graph $G=(V,E)$ and a function $c:V\mapsto\{1,2\}$. The function $c(\cdot)$ divides the vertices into two disjoint sets $V_1$ and $V_2$, where for all $v_1\in V_1$, we have $c(v_1)=1$ and for all $v_2\in V_2$, we have $c(v_2)=2$. There is no edge between $v_2\in V_2$ and $v_2'\in V_2$. There are edges between $v_1\in V_1$ and $v_2\in V_2$ and between $v_1\in V_1$ and $v_1'\in V_1$ (for $v_1\ne v_1'$). Find a two-to-one matching between $V_1$ and $V_2$ of maximum cardinality. That is, a set $S:=S_1\times S_2\subset V_1\times V_2$ such that:
- cardinality($S$) is maximum;
- for each $(s_1,s_2)\in S$, $\{s_1,s_2\}\in E$ ;
- $S_1$ is an independent set;
- for each $s_2\in S_2$, degree($s_2$) is at most $2$; and
- for each $s_1\in S_1$, degree($s_1$) is at most $1$. $\leftarrow$ EDIT
Is this problem NP-hard?
I was trying to prove that this problem is NP-hard by reduction from Conflict-Aware Weighted Bipartite B-Matching Problem (CA-WBM).
The authors proved that CA-WBM is NP-hard by a reduction from weighted independent set problem. Since I do not have weights in my problem, I was able to prove that even the Conflict-Aware Unweighted Bipartite B-Matching Problem (CA-UBM) is NP-hard (since the unweighted independent set is NP-hard). But I could not reduce CA-UBM to my problem. The issue is that in my problem the cost function is restricted to take only two values $\{1,2\}$.