Abstract
Braess’s paradox states that removing a part of a network may improve the players’ latency at equilibrium. In this work, we study the approximability of the best subnetwork problem for the class of random \({\mathcal {G}}_{n,p}\) instances proven prone to Braess’s paradox by Valiant and Roughgarden RSA ’10 (Random Struct Algorithms 37(4):495–515, 2010), Chung and Young WINE ’10 (LNCS 6484:194–208, 2010) and Chung et al. RSA ’12 (Random Struct Algorithms 41(4):451–468, 2012). Our main contribution is a polynomial-time approximation-preserving reduction of the best subnetwork problem for such instances to the corresponding problem in a simplified network where all neighbors of source s and destination t are directly connected by 0 latency edges. Building on this, we consider two cases, either when the total rate r is sufficiently low, or, when r is sufficiently high. In the first case of low \(r= O(n_{+})\), here \(n_{+}\) is the maximum degree of \(\{s, t\}\), we obtain an approximation scheme that for any constant \(\varepsilon > 0\) and with high probability, computes a subnetwork and an \(\varepsilon \)-Nash flow with maximum latency at most \((1+\varepsilon )L^*+ \varepsilon \), where \(L^*\) is the equilibrium latency of the best subnetwork. Our approximation scheme runs in polynomial time if the random network has average degree \(O(\mathrm {poly}(\ln n))\) and the traffic rate is \(O(\mathrm {poly}(\ln \ln n))\), and in quasipolynomial time for average degrees up to o(n) and traffic rates of \(O(\mathrm {poly}(\ln n))\). Finally, in the second case of high \(r= {\varOmega }(n_{+})\), we compute in strongly polynomial time a subnetwork and an \(\varepsilon \)-Nash flow with maximum latency at most \((1+2\varepsilon + o(1))L^*\).
Similar content being viewed by others
Notes
If the direction of a solution \(r_0\) gives increase to \(\ell _{q_{kl}}(f')\) then, because of linearity, the direction of \(-r_0\) decreases it. So we can assume that we can hit one of the constraints and have \(\ell _{q_{kl}}(f')\le L+b_{kl}\) (e.g. by choosing the direction that decreases \(\ell _{q_{kl}}(f')\)). Note that in both directions of \(r_0\) and \(-r_0\) we are bounded by one of the constraints because, in any direction, either some \(r_{q}\) would be increasing and \(f_q-r_q\ge 0\) will be bounding us, or all \(r_q\)’s would be decreasing and \(f_{kl}+\sum _qr_q\ge 0\) will be bounding us.
Recall Sect. 1, paragraph Previous Work, that for the optimum cost, we compute the flow X that minimizes the total latency cost \(C(X)= \sum _{e \in E(H_0^*)} x_e (a_e \cdot x_e + b_e)\) incurred by all users on network \(H_0^*\) wrt flow X.
With probability at least \(1 - \mathrm {e}^{-\delta _{\varepsilon } n_{-}/8}\).
This holds because the flow travelling from u to v inside G faces no capacities. For this flow and the edges used by it in G, a new minimization problem similar to the one above could be defined. The objective would be the same and the flows other the one going from u to v would be handled as constants. A solution to this problem could be used to minimize the initial objective and that’s why the initial problem returns a solution that solves also the new problem, which in turn has an equilibrium flow as a solution.
We repeatedly use the following form of the Chernoff bound (see e.g., [13]): Let \(X_1, \ldots , X_k\) be random variables independently distributed in \(\{0, 1\}\), and let \(X = \sum _{i=1}^k X_i\). Then, for all \(\epsilon \in (0, 1)\), \(\mathrm {I\!P}[X < (1-\epsilon )\mathrm {I\!E}[X]] \le \mathrm {e}^{-\epsilon ^2\,\mathrm {I\!E}[X] / 2}\), where \(\mathrm {e}\) is the basis of natural logarithms.
wrt. the random choice of the latency functions of G.
References
Althöfer, I.: On sparse approximations to randomized strategies and convex combinations. Linear Algebra Appl. 99, 339–355 (1994)
Bollobás, B.: Random Graphs, 2nd Edition Cambridge Studies in Advanced Mathematics, vol. 3. Cambridge University Press, Cambridge (2001)
Braess, D.: Über ein paradox aus der Verkehrsplanung. Unternehmensforschung 12, 258–268 (1968)
Cheeseman, P., Kanefsky, B., Taylor, W.: Where the really hard problems are. IJCAI 1, 331–337 (1991)
Chung, F., Young, S.J.: Braess’s paradox in large sparse graphs. In: Proceedings of the 6th Workshop on Internet and Network Economics (WINE ’10), LNCS, vol. 6484, pp. 194–208 (2010)
Chung, F., Young, S.J., Zhao, W.: Braess’s paradox in expanders. Random Struct. Algorithms 41(4), 451–468 (2012)
Cole, R., Dodis, Y., Roughgarden, T.: How much can taxes help selfish routing? J. Comput. Syst. Sci. 72(3), 444–467 (2006)
Duffin, R.J.: Topology of serries-parallel networks. J. Math. Anal. Appl. 10, 303–318 (1965)
Erdos, P., Renyi, A.: On the evolution of random graphs. Publ. Math. Inst. Hung. Acad. Sci. 5, 17–61 (1960)
Fotakis, D., Kaporis, A.C., Spirakis, P.G.: Efficient methods for selfish network design. Theor. Comput. Sci. 448, 9–20 (2012)
Franco, J.: Results related to threshold phenomena research in satisfiability: lower bounds. Theor. Comput. Sci. 265, 147–157 (2001)
Friedman, E.: Genericity and congestion control in selfish routing. In: 43rd IEEE Conference on Decision and Control, pp. 4667–4672 (2004)
Hagerup, T., Rüb, C.: A guided tour of Chernoff bounds. Inf. Process. Lett. 33(6), 305–308 (1990)
Johnson, D.: The NP-completeness column: an ongoing guide. J. Algorithms 5(2), 284–299 (1984)
Johnson, D., Aragon, C., McGeoch, L., Shevon, C.: Optimization by simulated annealing: an experimental evaluation. Oper. Res. 37(6), 865–892 (1989)
Karp, R.: Combinatorics, complexity, and randomness (turing award lecture). Commun. ACM 29(2), 98–109 (1989)
Kelly, F.: The Princeton companion to mathematics. In: Gowers, T., Green, J., Leader, I. (eds.) The Mathematics of Traffic in Networks. Princeton University Press, Princeton (2008)
Kirkpatrick, S., Selman, B.: Critical behavior in the satisfiability of random Boolean expressions. Science 264, 1297–1301 (1994)
Lin, H.C., Roughgarden, T., Tardos, É., Walkover, A.: Stronger bounds on Braess’s paradox and the maximum latency of selfish routing. SIAM J. Discrete Math. 25(4), 1667–1686 (2011)
Lipton, R.J., Markakis, E., Mehta, A.: Playing large games using simple strategies. In: Proceedings of the 4th ACM Conference on Electronic Commerce (EC ’03), pp. 36–41 (2003)
Lipton, R.J., Young, N.E.: Simple strategies for large zero-sum games with applications to complexity theory. In: Proceedings of the 26th ACM symposium on theory of computing (STOC ’94), pp. 734–740 (1994)
Mézard, M., Parisi, G., Zecchina, R.: Analytic and algorithmic solution of random satisfiability problems. Science 297, 812–815 (2002)
Mitchell, D., Selman, B., Levesque, H.: Hard and easy distribution of SAT problems. In: Proceeings of 10th National Conference on Artificial Intelligence (AAAI ’92), pp. 459–465
Milchtaich, I.: Network topology and the efficiency of equilibrium. Games Econ. Behav. 57, 321–346 (2006)
Morris, P.: The breakthrough method for escaping local minima. AAAI 6, 40-45 (1993)
Murchland, J.D.: Braess’s paradox of traffic flow. Transp. Res. 4, 391–394 (1970)
Nagurney, A., Boyce, D.: Preface to “on a paradox of traffic planning”. Transp. Sci. 39(4), 443–445 (2005)
Pas, E.I., Principio, S.L.: Braess’s paradox: some new insights. Transp. Res. Part B 31(3), 265–276 (1997)
Patriksson, M.: The Traffic Assignment Problem—Models and Methods. Linköping Institute of Technology, Linköping (1991)
Roughgarden, T.: Selfish Routing. Ph.D dissertation, Cornell Univ., USA, May (2002). http://theory.stanford.edu/~tim/
Roughgarden, T.: Stackelberg scheduling strategies. SIAM J. Comput. 33(2), 332–350 (2004)
Roughgarden, T.: Selfish Routing and the Price of Anarchy. MIT press, Cambridge, MA (2005)
Roughgarden, T.: On the severity of braess’s paradox: designing networks for selfish users is hard. J. Comput. Syst. Sci. 72(5), 922–953 (2006)
Selman, B., Kautz, H., Cohen, B.: Local search strategies for satisfiability testing. DIMACS 26, 521–532 (1993)
Steinberg, R., Zangwill, W.I.: The prevalence of Braess’ paradox. Transp. Sci. 17(3), 301–318 (1983)
Valiant, G., Roughgarden, T.: Braess’s paradox in large random graphs. Random Struct. Algorithms 37(4), 495–515 (2010)
Végh, L.A.: Strongly polynomial algorithm for a class of minimum-cost flow problems with separable convex objectives. In: Proceedings of the 44th ACM Symposium on Theory of Computing (STOC ’12), pp. 27–40 (2012)
International Competition and Symposium on Satisfiability Testing, Beijing ’96, Beijing, China, March 15–17 (1996)
2nd Dimacs Implementation Challenge. In: Johnson, D., Trick, M. (eds.) Dimacs Series in Discrete Mathematics and TCS, 26(4), AMS (1996). http://dimacs.rutgers.edu/Volumes/Vol26.html
Author information
Authors and Affiliations
Corresponding author
Additional information
This work was supported by the project Algorithmic Game Theory, co-financed by the European Union (European Social Fund-ESF) and Greek national funds, through the Operational Program “Education and Lifelong Learning”, under the research funding program Thales, by the EU ERC project RIMACO, by EU ERC project ALGAME Grant Agreement no. 321171, and by the EU FP7/2007-13 (DG INFSO G4-ICT for Transport) project eCompass Grant Agreement no. 288094.
Rights and permissions
About this article
Cite this article
Fotakis, D., Kaporis, A.C., Lianeas, T. et al. Resolving Braess’s Paradox in Random Networks. Algorithmica 78, 788–818 (2017). https://doi.org/10.1007/s00453-016-0175-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00453-016-0175-2