TY - GEN
T1 - Learning pseudo-Boolean k-DNF and submodular functions
AU - Raskhodnikova, Sofya
AU - Yaroslavtsev, Grigory
PY - 2013
Y1 - 2013
N2 - We prove that any submodular function f : {0, 1}n → {0, 1, ..., k} can be represented as a pseudo-Boolean 2k-DNF formula. Pseudo-Boolean DNFs are a natural generalization of DNF representation for functions with integer range. Each term in such a formula has an associated integral constant. We show that an analog of Håstad's switching lemma holds for pseudo-Boolean k-DNFs if all constants associated with the terms of the formula are bounded. This allows us to generalize Mansour's PAC-learning algorithm for k-DNFs to pseudo-Boolean k-DNFs, and hence gives a PAC-learning algorithm with membership queries under the uniform distribution for submodular functions of the form f : {0, 1}n → {0, 1, ..., k}. Our algorithm runs in time polynomial in n, kO(k log k/∈) and log(1/δ) and works even in the agnostic setting. The line of previous work on learning submodular functions [Balcan, Harvey (STOC '11), Gupta, Hardt, Roth, Ullman (STOC '11), Cheraghchi, Klivans, Kothari, Lee (SODA '12)] implies only nO(k) query complexity for learning submodular functions in this setting, for fixed ∈ and δ. Our learning algorithm implies a property tester for submodularity of functions f : {0, 1}n → {0, . . . , k} with query complexity polynomial in n for k = O((log n/log log n)1/2) and constant proximity parameter ∈.
AB - We prove that any submodular function f : {0, 1}n → {0, 1, ..., k} can be represented as a pseudo-Boolean 2k-DNF formula. Pseudo-Boolean DNFs are a natural generalization of DNF representation for functions with integer range. Each term in such a formula has an associated integral constant. We show that an analog of Håstad's switching lemma holds for pseudo-Boolean k-DNFs if all constants associated with the terms of the formula are bounded. This allows us to generalize Mansour's PAC-learning algorithm for k-DNFs to pseudo-Boolean k-DNFs, and hence gives a PAC-learning algorithm with membership queries under the uniform distribution for submodular functions of the form f : {0, 1}n → {0, 1, ..., k}. Our algorithm runs in time polynomial in n, kO(k log k/∈) and log(1/δ) and works even in the agnostic setting. The line of previous work on learning submodular functions [Balcan, Harvey (STOC '11), Gupta, Hardt, Roth, Ullman (STOC '11), Cheraghchi, Klivans, Kothari, Lee (SODA '12)] implies only nO(k) query complexity for learning submodular functions in this setting, for fixed ∈ and δ. Our learning algorithm implies a property tester for submodularity of functions f : {0, 1}n → {0, . . . , k} with query complexity polynomial in n for k = O((log n/log log n)1/2) and constant proximity parameter ∈.
UR - http://www.scopus.com/inward/record.url?scp=84876038221&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84876038221&partnerID=8YFLogxK
U2 - 10.1137/1.9781611973105.98
DO - 10.1137/1.9781611973105.98
M3 - Conference contribution
AN - SCOPUS:84876038221
SN - 9781611972511
T3 - Proceedings of the Annual ACM-SIAM Symposium on Discrete Algorithms
SP - 1356
EP - 1368
BT - Proceedings of the 24th Annual ACM-SIAM Symposium on Discrete Algorithms, SODA 2013
PB - Association for Computing Machinery
T2 - 24th Annual ACM-SIAM Symposium on Discrete Algorithms, SODA 2013
Y2 - 6 January 2013 through 8 January 2013
ER -