Submodular Maximization Beyond Non-negativity: Guarantees, Fast Algorithms, and Applications

It is generally believed that submodular functions -- and the more general class of $\gamma$-weakly submodular functions -- may only be optimized under the non-negativity assumption $f(S) \geq 0$. In this paper, we show that once the function is expressed as the difference $f = g - c$, where $g$ is monotone, non-negative, and $\gamma$-weakly submodular and $c$ is non-negative modular, then strong approximation guarantees may be obtained. We present an algorithm for maximizing $g - c$ under a $k$-cardinality constraint which produces a random feasible set $S$ such that $\mathbb{E} \left[ g(S) - c(S) \right] \geq (1 - e^{-\gamma} - \epsilon) g(OPT) - c(OPT)$, whose running time is $O (\frac{n}{\epsilon} \log^2 \frac{1}{\epsilon})$, i.e., independent of $k$. We extend these results to the unconstrained setting by describing an algorithm with the same approximation guarantees and faster $O(\frac{n}{\epsilon} \log\frac{1}{\epsilon})$ runtime. The main techniques underlying our algorithms are two-fold: the use of a surrogate objective which varies the relative importance between $g$ and $c$ throughout the algorithm, and a geometric sweep over possible $\gamma$ values. Our algorithmic guarantees are complemented by a hardness result showing that no polynomial-time algorithm which accesses $g$ through a value oracle can do better. We empirically demonstrate the success of our algorithms by applying them to experimental design on the Boston Housing dataset and directed vertex cover on the Email EU dataset.

[1]  Olga Veksler,et al.  Fast approximate energy minimization via graph cuts , 2001, Proceedings of the Seventh IEEE International Conference on Computer Vision.

[2]  Maxim Sviridenko,et al.  A note on maximizing a submodular set function subject to a knapsack constraint , 2004, Oper. Res. Lett..

[3]  Alexandros G. Dimakis,et al.  Streaming Weak Submodularity: Interpreting Neural Networks on the Fly , 2017, NIPS.

[4]  Vasek Chvátal,et al.  The tail of the hypergeometric distribution , 1979, Discret. Math..

[5]  Niv Buchbinder,et al.  Constrained Submodular Maximization via a Non-symmetric Technique , 2016, Math. Oper. Res..

[6]  Chandra Chekuri,et al.  Submodular function maximization via the multilinear relaxation and contention resolution schemes , 2011, STOC '11.

[7]  Amin Karbasi,et al.  Greed Is Good: Near-Optimal Submodular Maximization via Greedy Optimization , 2017, COLT.

[8]  Hui Lin,et al.  A Class of Submodular Functions for Document Summarization , 2011, ACL.

[9]  Andreas Krause,et al.  Adaptive Submodularity: Theory and Applications in Active Learning and Stochastic Optimization , 2010, J. Artif. Intell. Res..

[10]  M. L. Fisher,et al.  An analysis of approximations for maximizing submodular set functions—I , 1978, Math. Program..

[11]  Laurence A. Wolsey,et al.  Best Algorithms for Approximating the Maximum of a Submodular Set Function , 1978, Math. Oper. Res..

[12]  Moran Feldman,et al.  Guess Free Maximization of Submodular and Linear Sums , 2018, Algorithmica.

[13]  Éva Tardos,et al.  Maximizing the Spread of Influence through a Social Network , 2015, Theory Comput..

[14]  Andreas Krause,et al.  Near-optimal Nonmyopic Value of Information in Graphical Models , 2005, UAI.

[15]  W. Hoeffding Probability Inequalities for sums of Bounded Random Variables , 1963 .

[16]  Andreas Krause,et al.  Lazier Than Lazy Greedy , 2014, AAAI.

[17]  Christos Faloutsos,et al.  Graph evolution: Densification and shrinking diameters , 2006, TKDD.

[18]  Jan Vondrák,et al.  Optimal approximation for submodular and supermodular optimization with bounded curvature , 2013, SODA.

[19]  Alejandro Ribeiro,et al.  Approximate Supermodularity Bounds for Experimental Design , 2017, NIPS.

[20]  M. Skala Hypergeometric tail inequalities: ending the insanity , 2013, 1311.5939.

[21]  Michel Minoux,et al.  Accelerated greedy algorithms for maximizing submodular set functions , 1978 .

[22]  Martial Hebert,et al.  Efficient Feature Group Sequencing for Anytime Linear Prediction , 2014, UAI.

[23]  Huy L. Nguyen,et al.  Constrained Submodular Maximization: Beyond 1/e , 2016, 2016 IEEE 57th Annual Symposium on Foundations of Computer Science (FOCS).

[24]  Abhimanyu Das,et al.  Submodular meets Spectral: Greedy Algorithms for Subset Selection, Sparse Approximation and Dictionary Selection , 2011, ICML.

[25]  Jeff A. Bilmes,et al.  Submodularity beyond submodular energies: Coupling edges in graph cuts , 2011, CVPR 2011.

[26]  Jan Vondrák,et al.  Submodular Maximization over Multiple Matroids via Generalized Exchange Properties , 2009, Math. Oper. Res..

[27]  Andreas Krause,et al.  Guarantees for Greedy Maximization of Non-submodular Functions with Applications , 2017, ICML.

[28]  Jure Leskovec,et al.  Local Higher-Order Graph Clustering , 2017, KDD.

[29]  Alexandros G. Dimakis,et al.  Scalable Greedy Feature Selection via Weak Submodularity , 2017, AISTATS.

[30]  My T. Thai,et al.  Fast Maximization of Non-Submodular, Monotonic Functions on the Integer Lattice , 2018, ICML.

[31]  Alexandros G. Dimakis,et al.  Restricted Strong Convexity Implies Weak Submodularity , 2016, The Annals of Statistics.

[32]  Jeff A. Bilmes,et al.  Using Document Summarization Techniques for Speech Data Subset Selection , 2013, NAACL.

[33]  D. Rubinfeld,et al.  Hedonic housing prices and the demand for clean air , 1978 .