Adaptive Submodularity: A New Approach to Active Learning and Stochastic Optimization

Solving stochastic optimization problems under partial observability, where one needs to adaptively make decisions with uncertain outcomes, is a fundamental but notoriously difficult challenge. In this paper, we introduce the concept of adaptive submodularity, generalizing submodular set functions to adaptive policies. We prove that if a problem satisfies this property, a simple adaptive greedy algorithm is guaranteed to be competitive with the optimal policy. We illustrate the usefulness of the concept by giving several examples of adaptive submodular objectives arising in diverse applications including sensor placement, viral marketing and pool-based active learning. Proving adaptive submodularity for these problems allows us to recover existing results in these applications as special cases and leads to natural generalizations.

[1]  Steven Skiena,et al.  Decision trees for geometric models , 1993, SCG '93.

[2]  Andreas Krause,et al.  Near-optimal Observation Selection using Submodular Functions , 2007, AAAI.

[3]  Jeff A. Bilmes,et al.  Average-Case Active Learning with Costs , 2009, ALT.

[4]  Ronald L. Graham,et al.  Performance bounds on the splitting algorithm for binary testing , 1974, Acta Informatica.

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

[6]  Jon Kleinberg,et al.  Maximizing the spread of influence through a social network , 2003, KDD '03.

[7]  Donald W. Loveland.  Performance bounds for binary testing with arbitrary weights , 2004, Acta Informatica.

[8]  Alexander Schrijver,et al.  Combinatorial optimization. Polyhedra and efficiency. , 2003 .

[9]  Jeff A. Bilmes,et al.  Interactive Submodular Set Cover , 2010, ICML.

[10]  Carsten Lund,et al.  On the hardness of computing the permanent of random matrices , 1996, STOC '92.

[11]  Andreas S. Schulz,et al.  Revisiting the Greedy Approach to Submodular Set Function Maximization , 2007 .

[12]  Satoru Fujishige,et al.  Submodular functions and optimization , 1991 .

[13]  Jan Vondrák,et al.  Stochastic Covering and Adaptivity , 2006, LATIN.

[14]  Andrew McCallum,et al.  Employing EM and Pool-Based Active Learning for Text Classification , 1998, ICML.

[15]  Sanjoy Dasgupta,et al.  Analysis of a greedy active learning strategy , 2004, NIPS.

[16]  Mukesh K. Mohania,et al.  Decision trees for entity identification: approximation algorithms and hardness results , 2007, PODS '07.

[17]  Robert D. Nowak,et al.  Noisy Generalized Binary Search , 2009, NIPS.

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

[19]  Amin Saberi,et al.  Stochastic Submodular Maximization , 2008, WINE.

[20]  Teresa M. Przytycka,et al.  On an Optimal Split Tree Problem , 1999, WADS.