The UAI Approximate Inference Competition is now over. For a summary of the competition click here
We are pleased to announce an upcoming evaluation of probabilistic approximate inference algorithms, as part of the UAI conference this July. All researchers working on inference in graphical models are encouraged to participate.
The evaluation will focus on the following computational tasks:
- Calculating MAP assignments
- Calculating partition functions
- Calculating marginals
Submitted algorithms can compete in some or all of these tasks.
New features in this year’s competition:
- A special track will evaluate how well inference algorithms do when used as a black box for parameter learning.
- There will be separate tracks for approximate algorithms that provide upper or lower bound guarantees and those that do not. Exact algorithms are of course invited to participate in any of these tracks.
- Participants will receive automated performance reports and will be able to submit improvements during the entire competition period. A summarized leader board will be updated continuously.
- There will be separate tracks for different running time regimes (i.e., 20 seconds, 20 minutes, two hours).
We look forward to your participation,
- Four free UAI registrations will be awarded to the winning participants
- The top participants will have 10 minutes to present their solvers in the main track of the conference
- May 3 – Dataset and submission format announced.
- May 10 – Submissions accepted on website.
- July 1 – Final deadline for submissions.
- July 4 – Final results and winners announced.
- July 8-11 – Results reported at the UAI conference.
- Gal Elidan – Hebrew University
- Amir Globerson – Hebrew University
- Jeff Bilmes – Univ. of Washington
- Rina Dechter – UC Irvine
- Peter Grunwald – CWI
- Isabelle Guyon – Clopinet
- Peter Spirtes – CMU
- Uri Heinemann – Hebrew University