Abstract
We study online learning of general convex sets and halfspaces on graphs. While online learning of halfspaces in Euclidean space is a classical learning problem, the corresponding problem on graphs is understudied. In this context, a set of vertices is convex if it contains all connecting shortest paths and a halfspace is a convex set whose complement is also convex. We discuss mistake bounds based on the Halving algorithm and shortest path covers. Halving achieves near-optimal bounds but is inefficient in general. The shortest path cover based algorithm is efficient but provides optimal bounds only for restricted graph families such as trees. To mitigate the weaknesses of both approaches, we propose a novel polynomial time algorithm which achieves near-optimal bounds on graphs that are \(K_{2,k}\) minor-free for some constant \(k\in \mathbb {N}\). In contrast to previous mistake bounds on graphs, which typically depend on the induced cut of the labelling, our bounds only depend on the graph itself. Finally, we discuss the agnostic version of this problem and introduce an adaptive variant of Halving for k-intersections of halfspaces.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Angluin, D.: Queries and concept learning. Mach. Learn. 2(4), 319–342 (1988)
Ben-David, S., Pál, D., Shalev-Shwartz, S.: Agnostic online learning. In: COLT (2009)
Bodlaender, H.L.: A linear-time algorithm for finding tree-decompositions of small treewidth. SIAM J. Comput. 25(6), 1305–1317 (1996)
Boley, M., Gärtner, T., Grosskreutz, H.: Formal concept sampling for counting and threshold-free local pattern mining. In: ICDM (2010)
Boley, M., Horváth, T., Poigné, A., Wrobel, S.: Listing closed sets of strongly accessible set systems with applications to data mining. Theoret. Comput. Sci. 411(3), 691–700 (2010)
Bressan, M., Cesa-Bianchi, N., Lattanzi, S., Paudice, A.: Exact recovery of clusters in finite metric spaces using oracle queries. In: COLT (2021)
Cesa-Bianchi, N., Gentile, C., Vitale, F.: Fast and optimal prediction on a labeled tree. In: COLT (2009)
Cesa-Bianchi, N., Gentile, C., Vitale, F., Zappella, G.: A correlation clustering approach to link classification in signed networks. In: COLT (2012)
Cesa-Bianchi, N., Gentile, C., Vitale, F., Zappella, G.: Random spanning trees and the prediction of weighted graphs. JMLR 14(1), 1251–1284 (2013)
Cesa-Bianchi, N., Lugosi, G.: Prediction, learning, and games (2006)
Coelho, E.M.M., Dourado, M.C., Sampaio, R.M.: Inapproximability results for graph convexity parameters. Theoret. Comput. Sci. 600, 49–58 (2015)
Csikós, M., Mustafa, N.H., Kupavskii, A.: Tight lower bounds on the VC-dimension of geometric set systems. JMLR 20(1), 2991–2998 (2019)
Diestel, R.: Graph Theory, 5th edn. Springer, Heidelberg (2017). https://doi.org/10.1007/978-3-662-53622-3
Duchet, P.: Convex sets in graphs, II. Minimal path convexity. J. Combinat. Theor. Ser. B 44(3), 307–316 (1988)
Farber, M., Jamison, R.E.: Convexity in Graphs and Hypergraphs. SIAM J. Algeb. Discrete Methods 7(3), 433–444 (1986)
Frances, M., Litman, A.: Optimal mistake bound learning is hard. Inf. Comput. 144(1), 66–82 (1998)
Ganter, B.: Random extents and random closure systems. In: CLA, vol. 959, pp. 309–318 (2011)
Gärtner, T., Garriga, G.C.: The cost of learning directed cuts. In: ECMLPKDD (2007)
Glantz, R., Meyerhenke, H.: On finding convex cuts in general, bipartite and plane graphs. Theoret. Comput. Sci. 695, 54–73 (2017)
Herbster, M., Lever, G., Pontil, M.: Online prediction on large diameter graphs. In: NIPS (2008)
Herbster, M., Pasteris, S., Ghosh, S.: Online prediction at the limit of zero temperature. In: NIPS (2015)
Herbster, M., Pontil, M.: Prediction on a graph with a perceptron. In: NIPS (2006)
Herbster, M., Pontil, M., Wainer, L.: Online learning over graphs. In: ICML (2005)
Horváth, T., Turán, G.: Learning logic programs with structured background knowledge. Artif. Intell. 128(1–2), 31–97 (2001)
Kearns, M.J., Vazirani, U.: An introduction to computational learning theory (1994)
Kleinberg, J.: Detecting a network failure. Int. Math. 1(1), 37–55 (2004)
Kranakis, E., Krizanc, D., Ruf, B., Urrutia, J., Woeginger, G.: The VC-dimension of set systems defined by graphs. Discret. Appl. Math. 77(3), 237–257 (1997)
Littlestone, N.: Learning quickly when irrelevant attributes abound: a new linear-threshold algorithm. Mach. Learn. 2(4), 285–318 (1988)
Littlestone, N., Warmuth, M.K.: The weighted majority algorithm. Inf. Comput. 108(2), 212–261 (1994)
Lu, H.I., Ravi, R.: Approximating maximum leaf spanning trees in almost linear time. J. Algorithms 29(1), 132–141 (1998)
Missura, O., Gärtner, T.: Predicting dynamic difficulty. In: NIPS (2011)
Novikoff, A.B.: On convergence proofs on perceptrons. In: Symposium on the Mathematical Theory of Automata (1962)
Pelayo, I.M.: Geodesic convexity in graphs. Springer, New York (2013). https://doi.org/10.1007/978-1-4614-8699-2
Rosenblatt, F.: The Perceptron: a probabilistic model for information storage and organization in the brain. Psychol. Rev. 65(6), 386 (1958)
Seiffarth, F., Horváth, T., Wrobel, S.: Maximal closed set and half-space separations in finite closure systems. In: ECMLPKDD (2019)
Stadtländer, E., Horváth, T., Wrobel, S.: Learning weakly convex sets in metric spaces. In: ECMLPKDD (2021)
Thiessen, M., Gärtner, T.: Active learning of convex halfspaces on graphs. In: NeurIPS (2021)
van de Vel, M.L.: Theory of convex structures (1993)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Thiessen, M., Gärtner, T. (2023). Online Learning of Convex Sets on Graphs. In: Amini, MR., Canu, S., Fischer, A., Guns, T., Kralj Novak, P., Tsoumakas, G. (eds) Machine Learning and Knowledge Discovery in Databases. ECML PKDD 2022. Lecture Notes in Computer Science(), vol 13716. Springer, Cham. https://doi.org/10.1007/978-3-031-26412-2_22
Download citation
DOI: https://doi.org/10.1007/978-3-031-26412-2_22
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-26411-5
Online ISBN: 978-3-031-26412-2
eBook Packages: Computer ScienceComputer Science (R0)