Skip to main content
Log in

A General Reduction Theorem with Applications to Pathwidth and the Complexity of MAX 2-CSP

  • Published:
Algorithmica Aims and scope Submit manuscript

Abstract

We prove a general reduction theorem which allows us to extend bounds for certain graph parameters on cubic graphs to bounds for general graphs taking into account the individual vertex degrees. As applications, we give an algorithm for Max \(2\)-CSP whose complexity matches the algorithm of Scott and Sorkin in the case of \(d\)-regular graphs, \(d \le 5\), but is otherwise faster. It also improves on the previously fastest known algorithm in terms of the average degree, given by Golovnev and Kutzkov. Also from the general theorem, we derive a bound for the pathwidth of a general graph which equals that of Fomin et al. and Gaspers for graphs of degree at most \(6\), but is smaller otherwise, and use this to give an improved exponential-space algorithm for Max \(2\)-CSP. Finally we use the general result to give a faster algorithm for Max \(2\)-CSP on claw-free graphs.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others

Notes

  1. There is also a technical report, due to Robson [26], which makes use of a detailed analysis done by computer to solve maximum independent set in \(O^{\star }(2^{.25n})\) time and exponential space.

References

  1. Arnborg, S., Proskurowski, A., Corneil, D.G.: Forbidden minors characterization of partial 3-trees. Discret. Math. 80, 1–19 (1990)

    Article  MATH  MathSciNet  Google Scholar 

  2. Bansal, N., Raman, V.: Upper bounds for MAXSAT: further improved, in Algorithms and computation Chennai. Lecture Notes in Computer Science, vol. 1741, pp. 247–258. Springer, Berlin (1999)

  3. Bodlaender, H.L.: A partial \(k\)-arboretum of graphs with bounded treewidth. Theor. Comput. Sci. 209, 1–45 (1998)

    Article  MATH  MathSciNet  Google Scholar 

  4. Cheng, C., McDermid, E., Suzuki, I.: Planarization and acyclic colorings of subcubic claw-free graphs. In: Kolman, P., Kratochvíl, J. (eds.) Graph-Theoretic Concepts in Computer Science, WG 2011 (Czech Republic, June 2011). Lecture Notes in Computer Science, vol. 6986, pp. 107–118. Springer, Berlin (2011)

  5. de Fraysseix, H., Ossona de Mendez, P.: A characterization of DFS cotree critical graphs. In: Mutzel, P., Jünger, M., Leipert, S. (eds.) Graph Drawing 2001 (Vienna, 23–26 Sept. 2001), Lecture Notes in Computer Science, vol. 2265, pp. 84–95. Springer, Berlin (2002)

  6. Della Croce, F., Kaminski, M.J., Paschos, V.T.H.: An exact algorithm for MAX-CUT in sparse graphs. Oper. Res. Lett. 35, 403–408 (2007)

    Article  MATH  MathSciNet  Google Scholar 

  7. Edwards, K.J., Farr, G.F.: Fragmentability of graphs. J. Combin. Theory 82, 30–37 (2001)

    Article  MATH  MathSciNet  Google Scholar 

  8. Edwards, K.J., Farr, G.E.: Improved upper bounds for planarization and series-parallelization of average degree bounded graphs. Electron. J. Comb. 19(2), #P25 (2010)

    MathSciNet  Google Scholar 

  9. Edwards, K.J., Farr, G.E.: Graph fragmentability. In: Beineke, L.W., Wilson, R.J. (eds.) Topics in Structural Graph Theory, Encyclopedia of Mathematics and its Applications No. 147, pp. 203–218, Cambridge University Press, Cambridge (2013). ISBN 978-0-521-80231-4

  10. Fomin, F.V., Gaspers, S., Saurabh, S., Stepanov, A.A.: On two techniques of combining branching and treewidth. Algorithmica 54, 181–207 (2009)

    Article  MATH  MathSciNet  Google Scholar 

  11. Fomin, F.V., Grandoni, F., Kratsch, D.: A measure & conquer approach for the analysis of exact algorithms. J. ACM 56, 1–32 (2009)

    Article  MathSciNet  Google Scholar 

  12. Fomin, F.V., Høie, K.: Pathwidth of cubic graphs and exact algorithms. Inf. Process. Lett. 97, 191–196 (2006)

    Article  MATH  Google Scholar 

  13. Fomin, F.V., Kratsch, D.: Exact Exponential Algorithms. Springer, Texts in Theoretical Computer Science (2010). ISBN 978-3-642-16533-7

  14. Gaspers, S.: Exponential Time Algorithms: Structures, Measures, and Bounds. VDM Verlag Dr. Mueller e.K., 2010. ISBN 978-3-639-21825-1

  15. Gaspers, S., Sorkin, G.B.: A universally fastest algorithm for Max 2-Sat, Max 2-CSP, and everything in between. J. Comput. Syst. Sci. 78, 305–335 (2012)

    Article  MATH  MathSciNet  Google Scholar 

  16. Golovnev, A.: New upper bounds for MAX-2-SAT and MAX-2-CSP w.r.t. the average variable degree. In: Parameterized and exact computation, Lecture Notes in Computer Science, vol. 7112, pp. 106–117. Springer, Heidelberg (1973)

  17. Golovnev, A., Kutzkov, K.: New exact algorithms for the 2-constraint satisfaction problem. Theor. Comput. Sci. 526, 18–27 (2014)

    Article  MATH  MathSciNet  Google Scholar 

  18. Gramm, J., Hirsch, E.A., Niedermeier, R., Rossmanith, P.: Worst-case upper bounds for MAX-2-SAT with an application to MAX-CUT. Discret. Appl. Math. 130, 139–155 (2003)

    Article  MATH  MathSciNet  Google Scholar 

  19. Haxell, P., Pikhurko, O., Thomason, A.: Maximum acyclic and fragmented sets in regular graphs. J. Graph Theory 57, 149–156 (2008)

    Article  MATH  MathSciNet  Google Scholar 

  20. Hirsch, E.A.: A new algorithm for MAX-2-SAT, in: STACS 2000 (Lille), Lecture Notes in Computer Science, vol. 1770, pp. 65–73. Springer, Berlin (2000)

  21. Kneis, J., Mölle, D., Richter, S., Rossmanith, P.: A bound on the pathwidth of sparse graphs with applications to exact algorithms. SIAM J. Discret. Math. 23, 407–427 (2009)

    Article  MATH  Google Scholar 

  22. Kojevnikov, A., Kulikov, A.S.: A new approach to proving upper bounds for MAX-2-SAT. In: Proceedings of the Seventeenth Annual ACM-SIAM Symposium on Discrete Algorithms 2006, pp. 11–17. ACM, New York (2006)

  23. Kulikov, A.S., Kutzkov, K.: New bounds for MAX-SAT by clause learning, in: Proceedings of the 2nd International Symposium on Computer Science in Russia (CSR 2007), Lecture Notes in Computer Science, vol. 4649, pp. 194–204. Springer (2007)

  24. Niedermeier, R., Rossmanith, P.: New upper bounds for maximum satisfiability. J. Algorithms 36, 63–88 (2000)

    Article  MATH  MathSciNet  Google Scholar 

  25. Raible, D., Fernau, H.: A new upper bound for Max-2-SAT: a graph-theoretic approach. In: Mathematical foundations of computer science 2008, Lecture Notes in Computer Science, vol. 5162, pp. 551–562. Springer, Berlin, (2008)

  26. Robson, J.M.: Finding a maximum independent set in time \(O(2^{n/4})\). Technical Report 1251–01, LaBRI, Université Bordeaux I (2001)

  27. Scott, A.D., Sorkin, G.B.: Linear-programming design and analysis of fast algorithms for Max 2-CSP. Discret. Optim. 4, 260–287 (2007)

    Article  MATH  MathSciNet  Google Scholar 

  28. Scott, A.D., Sorkin, G.B.: Polynomial constraint satisfaction problems, graph bisection, and the Ising partition functio. ACM Trans. Algorithms 5, 45 (2009)

    Article  MathSciNet  Google Scholar 

  29. Williams, R.: A new algorithm for optimal constraint satisfaction and its implications. In: J. Díaz et al. (eds.), Proc. 31st International Colloquium on Automata, Languages and Programming (ICALP) (Turku, Finland, 2004), Lecture Notes in Computer Science, vol. 3142, pp. 1227–1237. Springer (2004)

  30. Woeginger, G.J.: Exact algorithms for NP-hard problems: a survey. In: Combinatorial optimization–Eureka, you shrink!, Lecture Notes in Computer Science vol. 2570, pp. 185–207. Springer (2003)

  31. Woeginger, G.J.: Open problems around exact algorithms. Discret. Appl. Math. 156, 397–405 (2008)

    Article  MATH  MathSciNet  Google Scholar 

  32. Xiao, M., Nagamochi, H.: Exact algorithms for maximum independent set. In: L. Cai, S.-W. Cheng, and T.-W. Lam (eds.), ISAAC2013, Lecture Notes in Computer Science, vol. 8283, pp. 328–338. Springer (2013)

Download references

Acknowledgments

We would like to thank David Wood for reminding us of the connection between planarization and Max 2-CSP, and the anonymous referees for their helpful comments and for drawing to our attention some recent work which we were unaware of.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Keith Edwards.

Appendix

Appendix

In this section we prove the properties of the functions \(g_{\alpha }\) and \(g_{\alpha }'\) which are listed in Sect. 3.1. First recall the definition of the functions concerned:

For any \(n \ge 2\), and \(\alpha \) with \(0 \le \alpha \le 1\), define the function \(g_{\alpha }(n)\) by setting \(g_{\alpha }(2)=0\), \(g_{\alpha }(3)=\alpha \), and for any \(n \ge 4\),

$$\begin{aligned} n g_{\alpha }(n) = (n-2) g_{\alpha }(n-1) + g_{\alpha }(n-2) + 1. \end{aligned}$$
(6)

We extend \(g_{\alpha }\) to all real numbers at least \(2\) by linear interpolation, i.e., if \(r = n + x\), where \(n \ge 2\) is an integer, and \(0 \le x \le 1\), then we set \(g_{\alpha }(r) = (1-x)g_{\alpha }(n) + xg_{\alpha }(n+1)\).

Also, for any \(n \ge 3\), define \(g_{\alpha }'(n) = g_{\alpha }(n) - g_{\alpha }(n-1)\).

Lemma 7.1

For any integer \(n \ge 2\),

$$\begin{aligned} g_{\alpha }(n) = (4-3\alpha ) \frac{A(n)}{n!} + (2-3\alpha )\frac{(-1)^n}{n!} - (3-3\alpha ) \end{aligned}$$

where \(A(n)\) is the alternating factorial function given by

$$\begin{aligned} A(n) = n! - (n-1)! + \cdots - (-1)^n \cdot 1!. \end{aligned}$$

Proof

Let \(h(n) = 1-g_{\alpha }(n)\). Then from the recurrence (6) above we obtain, for \(n \ge 4\),

$$\begin{aligned} n(1-h(n)) = (n-2)(1-h(n-1)) + 1 - h(n-2) + 1 \end{aligned}$$

or

$$\begin{aligned} nh(n) = (n-2)h(n-1) + h(n-2), \end{aligned}$$

from which

$$\begin{aligned} nh(n) + h(n-1) = (n-1)h(n-1) + h(n-2) \end{aligned}$$

Thus \(nh(n) + h(n-1)\) is a constant, \(K\) say, for all \(n \ge 3\). Multiplying through by \((n-1)!\), we obtain

$$\begin{aligned} n!h(n) + (n-1)!h(n-1) = K(n-1)! \end{aligned}$$

or (replacing \(n\) by \(n+1\))

$$\begin{aligned} (n+1)!h(n+1) + n!h(n) = Kn! \end{aligned}$$

for \(n \ge 2\). Setting \(j(n) = (n+1)!h(n+1)/K\) for \(n \ge 1\), we obtain

$$\begin{aligned} j(n) + j(n-1) = n! \end{aligned}$$

for all \(n \ge 2\). The alternating factorial function \(A(n)\) satisfies

$$\begin{aligned} A(n) + A(n-1) = n! \end{aligned}$$

so that

$$\begin{aligned} j(n) - A(n) = -(j(n-1) - A(n-1)). \end{aligned}$$

Hence \(j(n) - A(n) = C(-1)^n\) for some constant \(C\). Thus

$$\begin{aligned} h(n) = (K/n!)j(n-1) = (K/n!)A(n-1) - (KC/n!)(-1)^n \end{aligned}$$

or

$$\begin{aligned} g_{\alpha }(n) = 1 - (K/n!)(n! - A(n)) + (KC/n!)(-1)^n. \end{aligned}$$

Thus

$$\begin{aligned} g_{\alpha }(n) = (K/n!)A(n) + (KC/n!)(-1)^n - (K-1). \end{aligned}$$

Recall that \(K = (n+1)h(n+1) + h(n)\), so setting \(n=2\) gives \(K = 3(1-\alpha )+1 = 4 - 3\alpha \). Also \(C = A(1) - j(1) = 1 - 2h(2)/K\), so \(KC = K - 2h(2) = 4 - 3\alpha -2 = 2 - 3\alpha \). So finally,

$$\begin{aligned} g_{\alpha }(n) = (4-3\alpha ) \frac{A(n)}{n!} + (2-3\alpha )\frac{(-1)^n}{n!} - (3-3\alpha ) \end{aligned}$$

as required. \(\square \)

Lemma 7.2

For all real \(d \ge 2\),

$$\begin{aligned} g_{\alpha }(d) = 1 - \frac{4-3\alpha }{d+1} + O(1/d^3). \end{aligned}$$

Proof

We have, for \(n \ge 2\),

$$\begin{aligned} g_{\alpha }(n)&= (4-3\alpha ) \frac{A(n)}{n!} + (2-3\alpha )\frac{(-1)^n}{n!} -(3-3\alpha )\\&= (4\!-\!3\alpha ) \frac{n!-(n-1)!+(n-2)!-A(n-3)}{n!} \!+\! (2-3\alpha )\frac{(-1)^n}{n!} -(3-3\alpha )\\&= (4-3\alpha ) \frac{n!-(n-1)!+(n-2)!}{n!} - (3-3\alpha ) + O(1/n^3)\\&= 1 - (4-3\alpha ) \left( \frac{1}{n} - \frac{1}{n(n-1)} \right) + O(1/n^3)\\&= 1 - \frac{(4-3\alpha )}{n+1} + (4-3\alpha ) \left( \frac{1}{n+1} - \frac{1}{n} + \frac{1}{n(n-1)} \right) + O(1/n^3)\\&= 1 - \frac{(4-3\alpha )}{n+1} + (4-3\alpha ) \left( \frac{2}{n(n-1)(n+1)} \right) + O(1/n^3)\\&= 1 - \frac{(4-3\alpha )}{n+1} + O(1/n^3). \end{aligned}$$

The extension to non-integer arguments is straightforward. \(\square \)

Lemma 7.3

If \(\alpha \le 1/2\), then \(g_{\alpha }\) is non-decreasing, i.e.,

$$\begin{aligned} g_{\alpha }(n+1) \ge g_{\alpha }(n) \end{aligned}$$

for all integers \(n \ge 2\).

Proof

This is proved by induction, exactly as for the function \(g = g_{1/4}\) in [8]. First a very easy induction shows that \(g_{\alpha } < 1\) for all \(n\). Then we show by induction that for all \(n \ge 3\),

$$\begin{aligned} g_{\alpha }(n-1) \le g_{\alpha }(n) \le (1 + g_{\alpha }(n-1))/2 \end{aligned}$$

This is true for \(n = 3\), so suppose it is true for some \(n\). Then firstly,

$$\begin{aligned} (n+1)g_{\alpha }(n+1)&= (n-1)g_{\alpha }(n) + g_{\alpha }(n-1) + 1 \ge (n-1)g_{\alpha }(n) + 2g_{\alpha }(n) \\&= (n+1)g_{\alpha }(n), \end{aligned}$$

and secondly

$$\begin{aligned} (n+1)g_{\alpha }(n+1) = (n-1)g_{\alpha }(n) + g_{\alpha }(n-1) + 1 \le n g_{\alpha }(n) + 1 \end{aligned}$$

so that

$$\begin{aligned} g_{\alpha }(n+1) \le \frac{n}{n+1} g_{\alpha }(n) + \frac{1}{n+1} \le \frac{1}{2} g_{\alpha }(n) + \frac{1}{2} \end{aligned}$$

since \(n \ge 1\) and \(g_{\alpha }(n) < 1\). \(\square \)

Note that if \(\alpha > 1/2, g_{\alpha }\) is not non-decreasing, since \(g_{\alpha }(4) = (1 + 2\alpha )/4 < \alpha = g_{\alpha }(3)\).

Lemma 7.4

If \(1/6 \le \alpha \le 3/10\), then \(g_{\alpha }'\) is non-increasing, i.e.,

$$\begin{aligned} g_{\alpha }'(n+1) \le g_{\alpha }'(n) \end{aligned}$$

for all integers \(n \ge 3\).

Proof

We have

$$\begin{aligned} n g_{\alpha }(n)&= (n-2) g_{\alpha }(n-1) + g_{\alpha }(n-2) + 1, \\ (n-1) g_{\alpha }(n-1)&= (n-3) g_{\alpha }(n-2) + g_{\alpha }(n-3) + 1 \end{aligned}$$

from which we obtain

$$\begin{aligned} n g_{\alpha }'(n) = (n-3)g_{\alpha }'(n-1) + g_{\alpha }'(n-2). \end{aligned}$$

for \(n \ge 5\). Then it is easily shown by an induction similar to the one above that

$$\begin{aligned} g_{\alpha }'(n-1) \ge g_{\alpha }'(n) \ge g_{\alpha }'(n-1)/3. \end{aligned}$$

for all \(n \ge 4\). The base case requires that \(g_{\alpha }'(3) \ge g_{\alpha }'(4) \ge g_{\alpha }'(3)/3\), or \(\alpha \ge (1-2\alpha )/4 \ge \alpha /3\), which is true since \(1/6 \le \alpha \le 3/10\). \(\square \)

Lemma 7.5

If \(1/6 \le \alpha \le 3/10, g_{\alpha }(n)/n\) is strictly decreasing for \(n \ge 5\).

Proof

First note that \(g_{\alpha }(6) = (50\alpha +53)/120 > 1/2\). Now \(g_{\alpha }(n+1)/(n+1) < g_{\alpha }(n)/n\) if and only if \(ng_{\alpha }(n+1) < (n+1)g_{\alpha }(n)\). Since

$$\begin{aligned} (n+1)g_{\alpha }(n+1) = (n-1)g_{\alpha }(n) + g_{\alpha }(n-1) + 1, \end{aligned}$$

we have

$$\begin{aligned} ng_{\alpha }(n+1) = (n-1)g_{\alpha }(n) + g_{\alpha }(n-1) - g_{\alpha }(n+1) + 1, \end{aligned}$$

so \(ng_{\alpha }(n+1) < (n+1)g_{\alpha }(n)\) if and only if

$$\begin{aligned} g_{\alpha }(n-1) - g_{\alpha }(n+1) + 1 < 2g_{\alpha }(n), \end{aligned}$$

or

$$\begin{aligned} 2g_{\alpha }(n+1) - g_{\alpha }'(n+1) + g_{\alpha }'(n) > 1, \end{aligned}$$

which is true since \(g_{\alpha }(n+1) \ge g_{\alpha }(6) > 1/2\) and \(g_{\alpha }'\) is non-increasing. In fact \(g_{\alpha }(n)/n\) attains its maximum at \(n=5\) for all \(\alpha \) in this range. \(\square \)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Edwards, K., McDermid, E. A General Reduction Theorem with Applications to Pathwidth and the Complexity of MAX 2-CSP. Algorithmica 72, 940–968 (2015). https://doi.org/10.1007/s00453-014-9883-7

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00453-014-9883-7

Keywords

Navigation