Beating the Random Ordering is Hard: Inapproximability of Maximum Acyclic Subgraph

We prove that approximating the max. acyclic subgraph problem within a factor better than 1/2 is unique games hard. Specifically, for every constant epsiv > 0 the following holds: given a directed graph G that has an acyclic subgraph consisting of a fraction (1-epsiv) of its edges, if one can efficiently find an acyclic subgraph of G with more than (1/2 + epsiv) of its edges, then the UGC is false. Note that it is trivial to find an acyclic subgraph with 1/2 the edges, by taking either the forward or backward edges in an arbitrary ordering of the vertices of G. The existence of a rho-approximation algorithmfor rho > 1/2 has been a basic open problem for a while. Our result is the first tight inapproximability result for an ordering problem. The starting point of our reduction isa directed acyclic subgraph (DAG) in which every cut isnearly-balanced in the sense that the number of forward and backward edges crossing the cut are nearly equal; such DAGs were constructed by Charikar et al. Using this, we are able to study max. acyclic subgraph, which is a constraint satisfaction problem (CSP) over an unbounded domain, by relating it to a proxy CSP over a bounded domain. The latter is then amenable to powerful techniques based on the invariance principle. Our results also give a super-constant factor inapproximability result for the feedback arc set problem. Using our reductions, we also obtain SDP integrality gapsfor both the problems.

[1]  Elchanan Mossel Gaussian Bounds for Noise Correlation of Functions , 2007, FOCS 2007.

[2]  Alantha Newman,et al.  Approximating the Maximum Acyclic Subgraph , 2000 .

[3]  Joseph Naor,et al.  Approximating Minimum Feedback Sets and Multicuts in Directed Graphs , 1998, Algorithmica.

[4]  Madhu Sudan,et al.  Sparse Random Linear Codes are Locally Decodable and Testable , 2007, 48th Annual IEEE Symposium on Foundations of Computer Science (FOCS'07).

[5]  Prasad Raghavendra,et al.  Optimal algorithms and inapproximability results for every CSP? , 2008, STOC.

[6]  Elchanan Mossel,et al.  Noise stability of functions with low influences: Invariance and optimality , 2005, IEEE Annual Symposium on Foundations of Computer Science.

[7]  Mihalis Yannakakis,et al.  Optimization, approximation, and complexity classes , 1991, STOC '88.

[8]  Johan Håstad Every 2-CSP allows nontrivial approximation , 2005, STOC '05.

[9]  Vijaya Ramachandran,et al.  Finding a Minimum Feedback Arc Set in Reducible Flow Graphs , 1988, J. Algorithms.

[10]  Raymond E. Miller,et al.  Complexity of Computer Computations , 1972 .

[11]  Moses Charikar,et al.  On the Advantage over Random for Maximum Acyclic Subgraph , 2007, 48th Annual IEEE Symposium on Foundations of Computer Science (FOCS'07).

[12]  Nisheeth K. Vishnoi,et al.  The Unique Games Conjecture, Integrality Gap for Cut Problems and Embeddability of Negative Type Metrics into l1 , 2005, FOCS.

[13]  C. Lucchesi,et al.  A Minimax Theorem for Directed Graphs , 1978 .

[14]  Bruce A. Reed,et al.  Packing directed circuits , 1996, Comb..

[15]  Paul D. Seymour,et al.  Packing directed circuits fractionally , 1995, Comb..

[16]  Bonnie Berger,et al.  Tight Bounds for the Maximum Acyclic Subgraph Problem , 1997, J. Algorithms.

[17]  Subhash Khot On the power of unique 2-prover 1-round games , 2002, STOC '02.

[18]  Subhash Khot,et al.  Vertex cover might be hard to approximate to within 2-epsilon , 2008, J. Comput. Syst. Sci..

[19]  Alantha Newman Cuts and Orderings: On Semidefinite Relaxations for the Linear Ordering Problem , 2004, APPROX-RANDOM.

[20]  Prasad Raghavendra,et al.  Sdp gaps and ugc hardness for multiway cut, 0-extension, and metric labeling , 2008, STOC.