D3: Data-driven disjunctive abstraction

Hila Peleg*, Sharon Shoham, Eran Yahav

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


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.

Original languageEnglish
Title of host publicationVerification, Model Checking, and Abstract Interpretation - 17th International Conference, VMCAI 2016, Proceedings
EditorsK. Rustan, M. Leino, Barbara Jobstmann
PublisherSpringer Verlag
Number of pages21
ISBN (Print)9783662491218
StatePublished - 2016
Externally publishedYes
Event17th International Conference on Verification, Model Checking, and Abstract Interpretation, VMCAI 2016 - St. Petersburg, United States
Duration: 17 Jan 201619 Jan 2016

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference17th International Conference on Verification, Model Checking, and Abstract Interpretation, VMCAI 2016
Country/TerritoryUnited States
CitySt. Petersburg


Dive into the research topics of 'D3: Data-driven disjunctive abstraction'. Together they form a unique fingerprint.

Cite this