SYNT 2015: 4th Workshop on Synthesis (with CAV), 18 July 2015 - Paper Abstracts

SYNT 2015 program will include:

Sumit Gulwani: Programming by Examples applied to Data Manipulation

Programming by Examples (PBE) involves synthesizing intended programs in an underlying domain-specific language from example based specifications (Espec). In this talk, I will formalize our notion of Espec and describe some principles behind designing appropriate domain-specific languages. A key technical challenge in PBE is to search for programs that are consistent with the Espec provided by the user. We have developed a divide-and-conquer based search paradigm that leverages deductive rules and version space algebras to achieve real time efficiency. Another technical challenge in PBE is to resolve the ambiguity that is inherent in the Espec. We use machine learning based ranking techniques to predict an intended program within a set of programs that are consistent with the Espec. We also leverage active-learning based user interaction models to help resolve ambiguity in the Espec. In this talk, I will demo few PBE technologies (FlashFill, FlashExtract, and FlashRelate) that have been developed using these principles for the domain of data manipulation. These technologies are useful for data scientists who typically spend 80% of their time cleaning data, and for 99% of those end users who do not know programming.

Aditya Nori: Probabilistic Programming: Algorithms, Implementation and Synthesis

Recent years have seen a huge shift in the kind of programs that most programmers write. Programs are increasingly data driven instead of being algorithm driven. They use various forms of machine learning techniques to build models from data, for the purpose of decision making. Indeed, search engines, social networks, speech recognition, computer vision, and applications that use data from clinical trials, biological experiments, and sensors, are all examples of data driven programs.

We use the term "probabilistic programs" to refer to data driven programs that are written using higher-level abstractions. Though they span various application domains, all data driven programs have to deal with uncertainty in the data, and face similar issues in design, debugging, optimization and deployment.

In this talk, we describe connections this research area called "Probabilistic Programming" has with programming languages and software engineering --- this includes language design, static and dynamic analysis of programs, and program synthesis. We survey current state of the art and speculate on promising directions for future research.

SYNT 2015 Main Page