results for au:Feder_T in:cs

- Correspondence homomorphisms are both a generalization of standard homomorphisms and a generalization of correspondence colourings. For a fixed target graph $H$, the problem is to decide whether an input graph $G$, with each edge labeled by a pair of permutations of $V(H)$, admits a homomorphism to $H$ `corresponding' to the labels, in a sense explained below. We classify the complexity of this problem as a function of the fixed graph $H$. It turns out that there is dichotomy -- each of the problems is polynomial-time solvable or NP-complete. While most graphs $H$ yield NP-complete problems, there are interesting cases of graphs $H$ for which the problem is solved by Gaussian elimination. We also classify the complexity of the analogous correspondence \em list homomorphism problems, and also the complexity of a \em bipartite version of both problems. We emphasize the proofs for the case when $H$ is reflexive, but, for the record, we include a rough sketch of the remaining proofs in an Appendix.
- Update : An issue has been found in the correctness of our algorithm, and we are working to resolve the issue. Until a resolution is found, we retract our main claim that our approach gives a combinatorial solution to the CSP conjecture. We remain hopeful that we can resolve the issues. We thank Ross Willard for carefully checking the algorithm and pointing out the mistake in the version of this manuscript. We briefly explain one issue at the beginning of the text, and leave the rest of the manuscript intact for the moment . Ross Willard is posting a more involved description of a counter-example to the algorithm in the present manuscript. We have an updated manuscript that corrects some issues while still not arriving at a full solution; we will keep this private as long as unresolved issues remain. Previous abstract : We consider the problem of finding a homomorphism from an input digraph G to a fixed digraph H. We show that if H admits a weak-near-unanimity polymorphism $\phi$ then deciding whether G admits a homomorphism to H (HOM(H)) is polynomial time solvable. This confirms the conjecture of Maroti and McKenzie, and consequently implies the validity of the celebrated dichotomy conjecture due to Feder and Vardi. We transform the problem into an instance of the list homomorphism problem where initially all the lists are full (contain all the vertices of H). Then we use the polymorphism $\phi$ as a guide to reduce the lists to singleton lists, which yields a homomorphism if one exists.
- Matrix partition problems generalize a number of natural graph partition problems, and have been studied for several standard graph classes. We prove that each matrix partition problem has only finitely many minimal obstructions for split graphs. Previously such a result was only known for the class of cographs. (In particular, there are matrix partition problems which have infinitely many minimal chordal obstructions.) We provide (close) upper and lower bounds on the maximum size of a minimal split obstruction. This shows for the first time that some matrices have exponential-sized minimal obstructions of any kind (not necessarily split graphs). We also discuss matrix partitions for bipartite and co-bipartite graphs.
- Trigraph list homomorphism problems (also known as list matrix partition problems) have generated recent interest, partly because there are concrete problems that are not known to be polynomial time solvable or NP-complete. Thus while digraph list homomorphism problems enjoy dichotomy (each problem is NP-complete or polynomial time solvable), such dichotomy is not necessarily expected for trigraph list homomorphism problems. However, in this paper, we identify a large class of trigraphs for which list homomorphism problems do exhibit a dichotomy. They consist of trigraphs with a tree-like structure, and, in particular, include all trigraphs whose underlying graphs are trees. In fact, we show that for these tree-like trigraphs, the trigraph list homomorphism problem is polynomially equivalent to a related digraph list homomorphism problem. We also describe a few examples illustrating that our conditions defining tree-like trigraphs are not unnatural, as relaxing them may lead to harder problems.
- Motivated by recently discovered privacy attacks on social networks, we study the problem of anonymizing the underlying graph of interactions in a social network. We call a graph (k,l)-anonymous if for every node in the graph there exist at least k other nodes that share at least l of its neighbors. We consider two combinatorial problems arising from this notion of anonymity in graphs. More specifically, given an input graph we ask for the minimum number of edges to be added so that the graph becomes (k,l)-anonymous. We define two variants of this minimization problem and study their properties. We show that for certain values of k and l the problems are polynomial-time solvable, while for others they become NP-hard. Approximation algorithms for the latter cases are also given.