TY - GEN
T1 - D3
T2 - 17th International Conference on Verification, Model Checking, and Abstract Interpretation, VMCAI 2016
AU - Peleg, Hila
AU - Shoham, Sharon
AU - Yahav, Eran
N1 - Publisher Copyright:
© Springer-Verlag Berlin Heidelberg 2016.
PY - 2016
Y1 - 2016
N2 - We address the problem of computing an abstraction for a set of examples, which is precise enough to separate them from a set of counterexamples. The challenge is to find an over-approximation of the positive examples that does not represent any negative example. Conjunctive abstractions (e.g., convex numerical domains) and limited disjunctive abstractions, are often insufficient, as even the best such abstraction might include negative examples. One way to improve precision is to consider a general disjunctive abstraction. We present D3, a new algorithm for learning general disjunctive abstractions. Our algorithm is inspired by widely used machine-learning algorithms for obtaining a classifier from positive and negative examples. In contrast to these algorithms which cannot generalize from disjunctions, D3 obtains a disjunctive abstraction that minimizes the number of disjunctions. The result generalizes the positive examples as much as possible without representing any of the negative examples. We demonstrate the value of our algorithm by applying it to the problem of data- driven differential analysis, computing the abstract semantic difference between two programs. Our evaluation shows that D3 can be used to effectively learn precise differences between programs even when the difference requires a disjunctive representation.
AB - We address the problem of computing an abstraction for a set of examples, which is precise enough to separate them from a set of counterexamples. The challenge is to find an over-approximation of the positive examples that does not represent any negative example. Conjunctive abstractions (e.g., convex numerical domains) and limited disjunctive abstractions, are often insufficient, as even the best such abstraction might include negative examples. One way to improve precision is to consider a general disjunctive abstraction. We present D3, a new algorithm for learning general disjunctive abstractions. Our algorithm is inspired by widely used machine-learning algorithms for obtaining a classifier from positive and negative examples. In contrast to these algorithms which cannot generalize from disjunctions, D3 obtains a disjunctive abstraction that minimizes the number of disjunctions. The result generalizes the positive examples as much as possible without representing any of the negative examples. We demonstrate the value of our algorithm by applying it to the problem of data- driven differential analysis, computing the abstract semantic difference between two programs. Our evaluation shows that D3 can be used to effectively learn precise differences between programs even when the difference requires a disjunctive representation.
UR - http://www.scopus.com/inward/record.url?scp=84955274789&partnerID=8YFLogxK
U2 - 10.1007/978-3-662-49122-5_9
DO - 10.1007/978-3-662-49122-5_9
M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???
AN - SCOPUS:84955274789
SN - 9783662491218
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 185
EP - 205
BT - Verification, Model Checking, and Abstract Interpretation - 17th International Conference, VMCAI 2016, Proceedings
A2 - Rustan, K.
A2 - Leino, M.
A2 - Jobstmann, Barbara
PB - Springer Verlag
Y2 - 17 January 2016 through 19 January 2016
ER -