First SIGPLAN Workshop on Probabilistic and Approximate Computing

Co-located with PLDI 2014 in Edinburgh, Scotland.

June 13, 2014


Research is increasingly focusing on computing in the presence of approximation and inexactness, partly due to inexact data (for example, from sensors or from machine learning methods), and partly due to the performance and power benefits that arise from deliberate use of approximation. These methods require new approaches to every aspect of the hardware and software stack, ranging from new hardware to new algorithms to new languages and formal methods.

This workshop is an effort to bring together constituents from across these diverse areas to discuss challenges, opportunities, abstractions, and foundations. This research area has the additional exciting aspect that substantive research contributions often require diverse participation from research areas that include architecture, programming languages, machine learning, and distributed systems. Our goal is to bring together members of these diverse communities and build a shared understanding of concepts, applications, foundations, and systems.


Topics considered in-scope for the workshop include:

  • Mechanisms for approximation in hardware and software
  • Abstractions for approximation and uncertainty in programs (PL support)
  • Performance and efficiency improvements based on approximation
  • Domain-specific solutions using approximate computing
  • Domains in which approximation and noisy data is the norm, such as medical and sensor data
  • Incorporating results from machine learning as approximations in programs
  • Formal reasoning about programs with approximations
  • Important applications that allow approximations
  • Concurrency and approximation
  • Approximation and privacy/security
  • Compiler optimizations in the presence of approximate computing
  • Clean abstractions for describing and using machine-learning techniques in programs


Program Chairs:
Emery Berger, University of Massachusetts Amherst
Ben Zorn, Microsoft Research

Sriram Sankaranarayanan, University of Colorado, Boulder
Dan Grossman, University of Washington
Swarat Chaudhuri, Rice University
Subhasish Mitra, Stanford University
Ravi Nair, IBM Research
Kathleen Fisher, Tufts
Margaret Martonosi, Princeton
Luc De Raedt, TU Leuven
Willem Visser, University of Stellenbosch
Sasa Misailovic, MIT


How to Answer "Haven't We Done This Already?'", and Challenges/Opportunities in Approximate Computing
Kathryn McKinley (Microsoft Research) and Luis Ceze (University of Washington)
Programming Language and Compiler Support for Uncertainties
Eva Darulova and Viktor Kuncak (EPFL)
In Defense of Probabilistic Static Analysis
Benjamin Livshits and Shuvendu K. Lahiri (Microsoft Research)
Accuracy-Aware Program Transformations
Sasa Misailovic (MIT)
A Case for Runtime Coordination of Accuracy-aware Applications and Power-aware Systems
Henry Hoffmann (University of Chicago)
The relationship between probabilistic and approximate computing
Vikash Mansinghka (MIT)
Tabular: Probabilistic Inference from Excel
Andrew D Gordon (Microsoft Research and University of Edinburgh), Thore Graepel and Nicolas Rolland (Microsoft Research), Johannes Borgstroem (Uupsala University), Claudio Russo (Microsoft Research), and Marcin Szymczak (University of Edinburgh)
Towards a Universal Probabilistic Computer: Programming Models, Architectures, and Beyond
Biplab Deka (University of Illinois), Swarat Chaudhuri (Rice University), and Rakesh Kumar (University of Illinois)
Designing a MCMC library for Probabilistic Programming
Rob Zinkov and Chung-chieh Shan (Indiana University)
There's Something About Bayes: Effective Probabilistic Programming for the Rest of Us (Keynote ZIP)
James Bornholt, Todd Mytkowicz, and Kathryn S. McKinley (Microsoft Research)
Probabilistic Inference in PRISM
Taisuke Sato (Tokyo Institute of Technology)
Probabilistic Programming: Concepts and Challenges
Angelika Kimmig and Luc De Raedt (KU Leuven)
On the Probabilistic Symbolic Analysis of Programs
Antonio Filieri (Stuttgart) and Corina S. Pasareanu (Carnegie Mellon Silicon Valley, NASA Ames)
Scalable Synthesis
Brandon Lucia and Todd Mytkowicz (Microsoft Research)
Programming Abstractions for Approximate Computing
Michael Carbin (MIT)
Two Approximate-Programmability Birds, One Statistical-Inference Stone (Keynote ZIP)
Adrian Sampson (UW)
Optimizing Out Overcomputation (Keynote ZIP)
Eric Schkufza and Alex Aiken (Stanford)
Trading Functionality for Power within Applications
Melanie Kambadur and Martha Kim (Columbia)
Uncertainty Quantification in High Performance Computing
Florian Augustin and Youssef Marzouk (MIT)