WAX 2016 Program
Conference A, Georgia Tech Hotel and Conference Center
Sunday, April 3, 2016
9:00–9:15am: Opening and introductions
Be ready to say who you are, where you’re affiliated with, and why you’re at the workshop.
9:15–10:00am: Keynote by Matthai Philipose, Microsoft Research
Continuous Mobile Vision: Why and How to Perform Computer Vision on Video from a Wearable Device
10:00am–10:15am: Lightning talks 1
- Armin Alaghi, University of Washington: Approximate Computing for Energy-Harvesting Sensor Nodes
- Yuhao Zhu, UT Austin: The Human Processing Unit (HPU) as a New Approximate Computing Substrate
- Jongse Park, Georgia Tech: AxGames: Towards Crowdsourcing Quality Target Determination in Approximate Computing
- Djordje Jevdjic, University of Washington: CloudApp: Approximate Cloud Store for Multimedia
10:15–10:30am: Break
10:30–11:15am: Session 1, Architecture
Talks are 10 minutes each. There will be a 3-minute (short!) moment for questions immediately afterward; in the rest of the time, all the speakers will be available to answer more questions from attendees.
- Synthesis of Quality Configurable Systems
Seogoo Lee, Behzad Boroujerdian, Lizy K. John, and Andreas Gerstlauer (The University of Texas at Austin)
slides - Approximate Flash Storage: A Feasibility Study
Amir Rahmati, Matthew Hicks, and Atul Prakash (University of Michigan)
slides - Approximate Computing and Microfluidic Cooling for Enhanced Machine Learning
Hardik Sharma, William Wahby, Thomas Sarvey, Muhannad S. Bakir, and Hadi Esmailzadeh (Georgia Institute of Technology)
slides
11:15–noon: Session 2, Connections
- WoCMan: Harnessing the Wisdom of the Crowds for High-Quality Estimates
Daniel W. Barowy and Emery D. Berger (University of Massachusetts Amherst) and Daniel Goldstein and Siddharth Suri (Microsoft Research)
slides - Approximate and Noisy Computing: Connections to the Information-Theory World
Frederic Sala, Clayton Schoeny, and Lara Dolecek (UCLA) - The Case for Compulsory Approximation
Adrian Sampson (Cornell & Microsoft Research)
slides
noon–1:30pm: Lunch with discussion groups
We’ll select topics for discussion over lunch, assign groups to [topics][], and choose a scribe for each group. The scribe will be responsible for reporting out from the discussion after lunch.
1:30–1:45pm: Discussion report-outs
Each group’s scribe has a few minutes to lead discussion on the results from the lunch topic.
1:45–2pm: Lightning talks 2
- Mark Buckler, Cornell: Approximate Sensing: Bringing Approximation to the Source
- Yuhao Zhu, UT Austin: End-to-end Quality Control in Approximate Computing
- Luis Ceze, University of Washington: Disciplined Inconsistency
- Ben Zorn, Microsoft Research: We Need a Handle on Approximate Computing Now!
2–2:45pm: Keynote by Naveen Verma, Princeton
A Look at the Hardware: What Can Approximation Buy Us, and How Can We Cash In?
2:45–3pm: Break
3–3:45pm: Session 3, Mechanisms
- Overcoming the Data-flow Limit on Parallelism with Structural Approximation
Vignesh Balaji, Brandon Lucia, and Radu Marculescu (CMU)
slides - CPSA: Compute Precisely Store Approximately
Animesh Jain, Parker Hill, Michael A. Laurenzano, and Md E. Haque (University of Michigan, Ann Arbor), Muneeb Khan (Uppsala University), and Scott Mahlke, Lingjia Tang, and Jason Mars (University of Michigan, Ann Arbor) - DOT APPROX: Making a Case for Dynamic Online Training for Function Approximation Techniques
Aurangzeb and Rudolf Eigenmann (Purdue University)
3:45–4:30pm: Session 4, Accuracy
- Statistical Error Bounds for Data Parallel Applications
Parker Hill, Michael Laurenzano, Babak Zamirai, Mehrzad Samadi, Scott Mahlke, Jason Mars, and Lingjia Tang (University of Michigan)
slides - Approximating to the Last Bit
Thierry Moreau (University of Washington) and Adrian Sampson (Cornell & Microsoft Research) and Luis Ceze and Mark Oskin (University of Washington)
slides - Towards more Precision in Approximate Computing
Radha Venkatagiri, Abdulrahman Mahmoud, and Sarita V. Adve (University of Illinois at Urbana Champaign)
4:30–5:00pm and thereafter: Debate
Topics:
- Do we really need strict quality guarantees?
- Does approximation really need to be general?