Double TechTalk on probabilistic programming and empirical software engineering
Join us for a special event - the chance for a double TechTalk and networking. The first talk "An Introduction to Probabilistic Programming with HackPPL" is by Arun Kumar, developer at Facebook and
14.00-14.45 An Introduction to Probabilistic Programming with HackPPL by Arun Kumar, Facebook
15.00-15.45 Empirical Software Engineering in the 21st Century by Professor Prem Devanbu
16.00-17.00 Networking and refreshments
An Introduction to Probabilistic Programming with HackPPL by Arun Kumar, Facebook
At Facebook, we're interested in making machine learning more accessible to our engineers by giving them tools to perform probabilistic reasoning. To support this goal, we're building a universal probabilistic programming language (PPL), called HackPPL, within our server-side language, Hack.
This talk will introduce probabilistic programming, explain the changes we needed to make to Hack to support probabilistic programming and explore applications, both within and outside Facebook, where probabilistic programming offers advantages over mainstream machine learning approaches.
Bio: Arun Kumar is one of the members of the Probabilistic Programming Languages team at Facebook where he works predominantly on language design and developer tooling. He’s excited to be part of a team that combines two of his interests, machine learning
Empirical Software Engineering in the 21st Century by Prem Devanbu, Professor at University of California, Davis
The advent of copious, time-series data (e.g., source code, comments, bug reports, code reviews) and authorship meta-data (who-when-why, from version control) from open-source projects has driven a flowering of natural experiments in Empirical Software Engineering. Many questions, which had long remained a matter of dogma, could be subjected to empirical analysis. Do static analysis tools work? Does the programming language used make a difference? Does distributed development lead to more buggy code? Are open-source development teams well-organized, or chaotic? (The Cathedral vs. the Bazaar). Do experienced programmers make fewer mistakes? etc. This data has also enabled studies of code corpora, which led us to the discovery of Naturalness. This talk will review some highlights of this journey.
Bio: Prem Devanbu received his B.Tech from IIT Madras, and his