Submodular set function
In mathematics, a submodular set function (also known as a submodular function) is a set function that, informally, describes the relationship between a set of inputs and an output, where adding more of one input has a decreasing…
≈ 16 min read
Submodular set function. encyclopegia.org, snapshot enwiki-001366703148. /a/submodular-set-function
Names in other languages
No names in other languages are held for this entry yet.
Article bodies are English and are not machine-translated. A name in another language is not a translation of the article.
In mathematics, a submodular set function (also known as a submodular function) is a set function that, informally, describes the relationship between a set of inputs and an output, where adding more of one input has a decreasing additional benefit (diminishing returns). The natural diminishing returns property which makes them suitable for many applications, including approximation algorithms, game theory (as functions modeling user preferences) and electrical networks. Recently, submodular functions have also found utility in several real world problems in machine learning and artificial intelligence, including automatic summarization, multi-document summarization, feature selection, active learning, sensor placement, image collection summarization and many other domains.[1][2][3][4]
Definition
edit edit sourceIf Ω is a finite set, a submodular function is a set function f:2Ω→ℝ, where 2Ω denotes the power set of Ω, which satisfies one of the following equivalent conditions.[5]
- For every X,Y⊆Ω with X⊆Y and every x∈Ω∖Y we have that f(X∪{x})−f(X)≥f(Y∪{x})−f(Y).
- For every S,T⊆Ω we have that f(S)+f(T)≥f(S∪T)+f(S∩T).
- For every X⊆Ω and x1,x2∈Ω∖X such that x1≠x2 we have that f(X∪{x1})+f(X∪{x2})≥f(X∪{x1,x2})+f(X), or equivalently, f(X∪{x1})−f(X)≥f(X∪{x1,x2})−f(X∪{x2}).
A nonnegative submodular function is also a subadditive function, but a subadditive function need not be submodular. If Ω is not assumed finite, then the above conditions are not equivalent. In particular a function f defined by f(S)=1 if S is finite and f(S)=0 if S is infinite satisfies the first condition above, but the second condition fails when S and T are infinite sets with finite intersection.
Types and examples of submodular functions
edit edit sourceMonotone
edit edit sourceA set function f is monotone if for every T⊆S we have that f(T)≤f(S). Examples of monotone submodular functions include:
- Linear (Modular) functions
- Any function of the form f(S)=∑i∈Swi is called a linear function. Additionally if ∀i,wi≥0 then f is monotone.
- Budget-additive functions
- Any function of the form f(S)=min{B, ∑i∈Swi} for each wi≥0 and B≥0 is called budget additive.[6]
- Coverage functions
- Let Ω={E1,E2,…,En} be a collection of subsets of some ground set Ω′. The function f(S)=|⋃Ei∈SEi| for S⊆Ω is called a coverage function. This can be generalized by adding non-negative weights to the elements.
- Entropy
- Let Ω={X1,X2,…,Xn} be a set of random variables. Then for any S⊆Ω we have that H(S) is a submodular function, where H(S) is the entropy of the set of random variables S, a fact known as Shannon's inequality.[7] Further inequalities for the entropy function are known to hold, see entropic vector.
- Matroid rank functions
- Let Ω={e1,e2,…,en} be the ground set on which a matroid is defined. Then the rank function of the matroid is a submodular function.[8]
Non-monotone
edit edit sourceA submodular function that is not monotone is called non-monotone. In particular, a function is called non-monotone if it has the property that adding more elements to a set can decrease the value of the function. More formally, the function f is non-monotone if there are sets S,T in its domain s.t. S⊂T and f(S)>f(T).
Symmetric
edit edit sourceA non-monotone submodular function f is called symmetric if for every S⊆Ω we have that f(S)=f(Ω−S). Examples of symmetric non-monotone submodular functions include:
- Graph cuts
- Let Ω={v1,v2,…,vn} be the vertices of a graph. For any set of vertices S⊆Ω let f(S) denote the number of edges e=(u,v) such that u∈S and v∈Ω−S. This can be generalized by adding non-negative weights to the edges.
- Mutual information
- Let Ω={X1,X2,…,Xn} be a set of random variables. Then for any S⊆Ω we have that f(S)=I(S;Ω−S) is a submodular function, where I(S;Ω−S) is the mutual information.
Asymmetric
edit edit sourceA non-monotone submodular function which is not symmetric is called asymmetric.
- Directed cuts
- Let Ω={v1,v2,…,vn} be the vertices of a directed graph. For any set of vertices S⊆Ω let f(S) denote the number of edges e=(u,v) such that u∈S and v∈Ω−S. This can be generalized by adding non-negative weights to the directed edges.
Continuous extensions of submodular set functions
edit edit sourceOften, given a submodular set function that describes the values of various sets, we need to compute the values of fractional sets. For example: we know that the value of receiving house A and house B is V, and we want to know the value of receiving 40% of house A and 60% of house B. To this end, we need a continuous extension of the submodular set function.
Formally, a set function f:2Ω→ℝ with |Ω|=n can be represented as a function on {0,1}n, by associating each S⊆Ω with a binary vector xS∈{0,1}n such that xiS=1 when i∈S, and xiS=0 otherwise. A continuous extension of f is a continuous function F:[0,1]n→ℝ, that matches the value of f on x∈{0,1}n, i.e. F(xS)=f(S).
Several kinds of continuous extensions of submodular functions are commonly used, which are described below.
Lovász extension
edit edit sourceThis extension is named after mathematician László Lovász.[9] Consider any vector 𝐱={x1,x2,…,xn} such that each 0≤xi≤1. Then the Lovász extension is defined as
fL(𝐱)=𝔼(f({i|xi≥λ}))
where the expectation is over λ chosen from the uniform distribution on the interval [0,1]. The Lovász extension is a convex function if and only if f is a submodular function.
Multilinear extension
edit edit sourceConsider any vector 𝐱={x1,x2,…,xn} such that each 0≤xi≤1. Then the multilinear extension is defined as [10][11]F(𝐱)=∑S⊆Ωf(S)∏i∈Sxi∏i∉S(1−xi).
Intuitively, xi represents the probability that item i is chosen for the set. For every set S, the two inner products represent the probability that the chosen set is exactly S. Therefore, the sum represents the expected value of f for the set formed by choosing each item i at random with probability xi, independently of the other items.
Convex closure
edit edit sourceConsider any vector 𝐱={x1,x2,…,xn} such that each 0≤xi≤1. Then the convex closure is defined as f−(𝐱)=min(∑SαSf(S):∑SαS1S=𝐱,∑SαS=1,αS≥0).
The convex closure of any set function is convex over [0,1]n.
Concave closure
edit edit sourceConsider any vector 𝐱={x1,x2,…,xn} such that each 0≤xi≤1. Then the concave closure is defined as f+(𝐱)=max(∑SαSf(S):∑SαS1S=𝐱,∑SαS=1,αS≥0).
Relations between continuous extensions
edit edit sourceFor the extensions discussed above, it can be shown that f+(𝐱)≥F(𝐱)≥f−(𝐱)=fL(𝐱) when f is submodular.[12]
Properties
edit edit source- The class of submodular functions is closed under non-negative linear combinations. Consider any submodular function f1,f2,…,fk and non-negative numbers α1,α2,…,αk. Then the function g defined by g(S)=∑i=1kαifi(S) is submodular.
- For any submodular function f, the function defined by g(S)=f(Ω∖S) is submodular.
- The function g(S)=min(f(S),c), where c is a real number, is submodular whenever f is monotone submodular. More generally, g(S)=h(f(S)) is submodular, for any non decreasing concave function h.
- Consider a random process where a set T is chosen with each element in Ω being included in T independently with probability p. Then the following inequality is true 𝔼[f(T)]≥pf(Ω)+(1−p)f(∅) where ∅ is the empty set. More generally consider the following random process where a set S is constructed as follows. For each of 1≤i≤l,Ai⊆Ω construct Si by including each element in Ai independently into Si with probability pi. Furthermore let S=∪i=1lSi. Then the following inequality is true 𝔼[f(S)]≥∑R⊆[l]Πi∈RpiΠi∉R(1−pi)f(∪i∈RAi).[citation needed]
Optimization problems
edit edit sourceSubmodular functions have properties which are very similar to convex and concave functions. For this reason, an optimization problem which concerns optimizing a convex or concave function can also be described as the problem of maximizing or minimizing a submodular function subject to some constraints.
Submodular set function minimization
edit edit sourceThe hardness of minimizing a submodular set function depends on constraints imposed on the problem.
- The unconstrained problem of minimizing a submodular function is computable in polynomial time,[13][14] and even in strongly-polynomial time.[15][16] Computing the minimum cut in a graph is a special case of this minimization problem.
- The problem of minimizing a submodular function with a cardinality lower bound is NP-hard, with polynomial factor lower bounds on the approximation factor.[17][18]
Submodular set function maximization
edit edit sourceUnlike the case of minimization, maximizing a generic submodular function is NP-hard even in the unconstrained setting. Thus, most of the works in this field are concerned with polynomial-time approximation algorithms, including greedy algorithms or local search algorithms.
- The problem of maximizing a non-negative submodular function admits a 1/2 approximation algorithm.[19][20] Computing the maximum cut of a graph is a special case of this problem.
- The problem of maximizing a monotone submodular function subject to a cardinality constraint admits a 1−1/e approximation algorithm.[21][22] The maximum coverage problem is a special case of this problem.
- The problem of maximizing a monotone submodular function subject to a matroid constraint (which subsumes the case above) also admits a 1−1/e approximation algorithm.[23][24][25]
Many of these algorithms can be unified within a semi-differential based framework of algorithms.[18]
Related optimization problems
edit edit sourceApart from submodular minimization and maximization, there are several other natural optimization problems related to submodular functions.
- Minimizing the difference between two submodular functions[26] is not only NP hard, but also inapproximable.[27]
- Minimization/maximization of a submodular function subject to a submodular level set constraint (also known as submodular optimization subject to submodular cover or submodular knapsack constraint) admits bounded approximation guarantees.[28]
- Partitioning data based on a submodular function to maximize the average welfare is known as the submodular welfare problem, which also admits bounded approximation guarantees (see welfare maximization).
Applications
edit edit sourceSubmodular functions naturally occur in several real world applications, in economics, game theory, machine learning and computer vision[4][29] as well as general artificial intelligence[30]. Owing to the diminishing returns property, submodular functions naturally model costs of items, since there is often a larger discount, with an increase in the items one buys. Submodular functions model notions of complexity, similarity and cooperation when they appear in minimization problems. In maximization problems, on the other hand, they model notions of diversity, information and coverage.
See also
edit edit sourceCitations
edit edit source- ^ H. Lin and J. Bilmes, A Class of Submodular Functions for Document Summarization, ACL-2011. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ S. Tschiatschek, R. Iyer, H. Wei and J. Bilmes, Learning Mixtures of Submodular Functions for Image Collection Summarization, NIPS-2014. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ A. Krause and C. Guestrin, Near-optimal nonmyopic value of information in graphical models, UAI-2005. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ a b A. Krause and C. Guestrin, Beyond Convexity: Submodularity in Machine Learning, Tutorial at ICML-2008 Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ (Schrijver 2003, §44, p. 766) Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Buchbinder, Niv; Feldman, Moran (2018). "Submodular Functions Maximization Problems". In Gonzalez, Teofilo F. (ed.). Handbook of Approximation Algorithms and Metaheuristics, Second Edition: Methodologies and Traditional Applications. Chapman and Hall/CRC. doi:10.1201/9781351236423. ISBN 9781351236423. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ "Information Processing and Learning" (PDF). cmu. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Fujishige (2005) p.22 Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Lovász, L. (1983). "Submodular functions and convexity". Mathematical Programming the State of the Art. pp. 235–257. doi:10.1007/978-3-642-68874-4_10. ISBN 978-3-642-68876-8. S2CID 117358746. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Vondrak, Jan (2008-05-17). "Optimal approximation for the submodular welfare problem in the value oracle model". Proceedings of the fortieth annual ACM symposium on Theory of computing. STOC '08. New York, NY, USA: Association for Computing Machinery. pp. 67–74. doi:10.1145/1374376.1374389. ISBN 978-1-60558-047-0. S2CID 170510. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Calinescu, Gruia; Chekuri, Chandra; Pál, Martin; Vondrák, Jan (January 2011). "Maximizing a Monotone Submodular Function Subject to a Matroid Constraint". SIAM Journal on Computing. 40 (6): 1740–1766. doi:10.1137/080733991. ISSN 0097-5397. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Vondrák, Jan. "Polyhedral techniques in combinatorial optimization: Lecture 17" (PDF). Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Grötschel, M.; Lovasz, L.; Schrijver, A. (1981). "The ellipsoid method and its consequences in combinatorial optimization". Combinatorica. 1 (2): 169–197. doi:10.1007/BF02579273. hdl:10068/182482. S2CID 43787103. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Cunningham, W. H. (1985). "On submodular function minimization". Combinatorica. 5 (3): 185–192. doi:10.1007/BF02579361. S2CID 33192360. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Iwata, S.; Fleischer, L.; Fujishige, S. (2001). "A combinatorial strongly polynomial algorithm for minimizing submodular functions". J. ACM. 48 (4): 761–777. doi:10.1145/502090.502096. S2CID 888513. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Schrijver, A. (2000). "A combinatorial algorithm minimizing submodular functions in strongly polynomial time". J. Combin. Theory Ser. B. 80 (2): 346–355. doi:10.1006/jctb.2000.1989. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Z. Svitkina and L. Fleischer, Submodular approximation: Sampling-based algorithms and lower bounds, SIAM Journal on Computing (2011). Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ a b R. Iyer, S. Jegelka and J. Bilmes, Fast Semidifferential based submodular function optimization, Proc. ICML (2013). Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ U. Feige, V. Mirrokni and J. Vondrák, Maximizing non-monotone submodular functions, Proc. of 48th FOCS (2007), pp. 461–471. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ N. Buchbinder, M. Feldman, J. Naor and R. Schwartz, A tight linear time (1/2)-approximation for unconstrained submodular maximization, Proc. of 53rd FOCS (2012), pp. 649-658. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Nemhauser, George; Wolsey, L. A.; Fisher, M. L. (1978). "An analysis of approximations for maximizing submodular set functions I". Mathematical Programming. 14 (14): 265–294. doi:10.1007/BF01588971. S2CID 206800425. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Williamson, David P. "Bridging Continuous and Discrete Optimization: Lecture 23" (PDF). Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ G. Calinescu, C. Chekuri, M. Pál and J. Vondrák, Maximizing a submodular set function subject to a matroid constraint, SIAM J. Comp. 40:6 (2011), 1740-1766. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ M. Feldman, J. Naor and R. Schwartz, A unified continuous greedy algorithm for submodular maximization, Proc. of 52nd FOCS (2011). Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Y. Filmus, J. Ward, A tight combinatorial algorithm for submodular maximization subject to a matroid constraint, Proc. of 53rd FOCS (2012), pp. 659-668. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ M. Narasimhan and J. Bilmes, A submodular-supermodular procedure with applications to discriminative structure learning, In Proc. UAI (2005). Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ R. Iyer and J. Bilmes, Algorithms for Approximate Minimization of the Difference between Submodular Functions, In Proc. UAI (2012). Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ R. Iyer and J. Bilmes, Submodular Optimization Subject to Submodular Cover and Submodular Knapsack Constraints, In Advances of NIPS (2013). Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ J. Bilmes, Submodularity in Machine Learning Applications, Tutorial at AAAI-2015. Wikipedia revision 1366703148, retrieved 2026-11-22
- ^ Bilmes, Jeff (Jan 31, 2022). "Submodularity In Machine Learning and Artificial Intelligence". arXiv:2202.00132 [cs.LG]. Wikipedia revision 1366703148, retrieved 2026-11-22
References
edit edit source- Schrijver, Alexander (2003), Combinatorial Optimization, Springer, ISBN 3-540-44389-4
- Lee, Jon (2004), A First Course in Combinatorial Optimization, Cambridge University Press, ISBN 0-521-01012-8
- Fujishige, Satoru (2005), Submodular Functions and Optimization, Elsevier, ISBN 0-444-52086-4
- Narayanan, H. (1997), Submodular Functions and Electrical Networks, Elsevier, ISBN 0-444-82523-1
- Oxley, James G. (1992), Matroid theory, Oxford Science Publications, Oxford: Oxford University Press, ISBN 0-19-853563-5, Zbl 0784.05002
External links
edit edit source- http://www.cs.berkeley.edu/~stefje/references.html has a longer bibliography
- http://submodularity.org/ includes further material on the subject
Text derived from https://en.wikipedia.org/w/index.php?oldid=1366703148 under CC BY-SA 4.0; modifications have been made.