[Pl-seminar] 3/24 Seminar: Swarat Chaudhuri, Learning to Program and Debug, Automatically

Daniel Patterson dbp at ccs.neu.edu
Thu Mar 23 10:09:55 EDT 2017


Reminder - this is tomorrow!

On Sun, Mar 12, 2017 at 12:14 PM Daniel Patterson <dbp at ccs.neu.edu> wrote:

> NUPRL Seminar presents
>
> Swarat Chaudhuri
> Rice University
> Host: Amal Ahmed
>
> 12:00-1:30PM
> Friday, March 24, 2017
> Room 366 WVH (http://www.ccs.neu.edu/home/wand/directions.html)
>
> Learning to Program and Debug, Automatically
>
> Abstract:
>
> Automating programming and debugging are long-standing goals in computer
> science. In spite of significant progress in formal methods over the years,
> we remain very far from achieving these goals. For example, a freshman CS
> major will typically program circles around today's best program
> synthesizers. Debugging and verification tools rely on formal
> specifications, which are hard to provide in many important applications.
>
> Two critical components of the gap between human and machine programmers
> are that humans learn from experience, i.e., data, and can easily
> generalize from incomplete problem definitions. In this talk, I will
> present a general framework for formal methods, based on Bayesian
> statistical learning, that aims to eliminate these differences. In our
> framework, descriptions of programming tasks are seen to be "clues" towards
> a hidden (probabilistic) specification that fully defines the task. Large
> corpora of real-world programs are used to construct a statistical model
> that correlates specifications with the form and function of their
> implementations. The framework can be implemented in a variety of ways, but
> in particular, through a neural architecture called Bayesian variational
> encoder-decoders. Inferences made using the framework can be used to guide
> traditional algorithms for program synthesis and bug-finding.
>
> I will show that this data-driven approach can lead to giant leaps in the
> scope and performance of automated program synthesis and debugging
> algorithms. Specifically, I will give a demo of Bayou, a system for
> Bayesian inductive synthesis of Java programs that goes significantly
> beyond the state of the art in program synthesis. I will also describe
> Salento, a debugging system based on our framework that can find subtle
> violations of API contracts without any kind of specification.
>
> Bio:
>
> Swarat Chaudhuri is an Associate Professor of Computer Science at Rice
> University. His research lies at the interface of programming systems and
> artificial intelligence. Much of his recent work is on program synthesis,
> the problem of automatically generating computer programs from high-level
> specifications.
>
>
>
-------------- next part --------------
HTML attachment scrubbed and removed


More information about the pl-seminar mailing list