AAAI-17 Workshop on Symbolic Inference and Optimization (SymInfOpt-17)

Overview

The purpose of the workshop is to explore and promote symbolic approaches to probabilistic inference, numerical optimization and machine learning. The workshop will place a special emphasis on techniques for mixed discrete/continuous (hybrid) domains and techniques that can be extended to such domains.

Symbolic approaches enjoy a long and distinguished history in AI. While the last two decades have seen major advances in probabilistic modeling, data management, data fusion and data‐driven learning, much of this work assumes fairly low‐level representations that are tailored for a specific application. It is now recognized that formal languages, and their symbolic underpinnings, can enable descriptive clarity, re‐usability, and interpretability, thereby furthering the applicability and impact of AI technology.

Recently, there have been significant successes of formal representations and symbolic techniques for inference and optimization. In the area of probabilistic modeling, weighted model counting has emerged as a competitive and general paradigm, providing state‐of‐the‐art inference for graphical models, Markov Logic Networks, and probabilistic programming. In the area of planning, symbolic approaches have been shown to handle large state spaces by leveraging abstractions. In the area of verification, real‐time systems in robotics and even pacemakers have been successfully checked for safety and correctness using symbolic specifications. In the area of optimization and learning, symbolic approaches ranging from symbolic algebra to SMT to decision diagrams have enabled novel scalable solutions. 

Encouraged by these successes, the workshop aspires to bring together AI researchers from knowledge representation, machine learning, databases, verification and planning to bettٰer understand applications of symbolic methods to inference and optimization problems across all fields.

Schedule (Sunday, Feb 5, 2017, Hilton San Francisco, Union Square 15-16) 


8:45 - 9:00: Welcome and Opening Remarks

9:00 - 9:40: Keynote 1: Stefano Ermon, Stanford, "Fourier Representations in Probabilistic Inference"

9:40 - 10:20: Keynote 2: Guy Van den Broeck, UCLA, "Tractable Learning in Structured Probability Spaces"


10:20 - 10:50: Coffee Break


10:50 - 11:30: Keynote 3: Steven Diamond, Stanford University, "Convex Optimization with Abstract Linear Operators"

11:30 - 12:15: Poster Highlights (9 min each): Symbolic Numerical Methods and Planning

  • Taisuke Sato, "Embedding Tarskian semantics in vector spaces"
  • Martin Mladenov, Vaishak Belle and Kristian Kersting, "Towards Symbolic-Numerical Optimization"
  • Shamin Kinathil, Harold Soh and Scott Sanner, "Nonlinear Optimization and Symbolic Dynamic Programming for Parameterized Hybrid Markov Decision Processes"
  • Fumito Takeuchi, Masaaki Nishino, Norihito Yasuda, Takuya Akiba, Shin-Ichi Minato and Masaaki Nagata, "BDD-Constrained A* Search: A Fast Method for Solving Constrained DAG Shortest-Path Problems"
  • Wiktor Piotrowski, Maria Fox, Derek Long, Daniele Magazzeni and Fabio Mercorio, "PDDL+ Planning with Temporal Pattern Databases"

12:15 - 14:00: Lunch Break


14:00 - 14:45: Poster Highlights (9 min each): Hybrid Logical Models and Machine Learning

  • Vaishak Belle, "Open-Universe Weighted Model Counting" (Extended Abstract)
  • Timothy Kopp, Parag Singla and Henry Kautz, "Conditional Term Equivalent Symmetry Breaking for SAT"
  • Cristina Cornelio and Vijay Saraswat, "Expressing Probabilistic Graphical Models in RCC"
  • Alejandro Molina, Sriraam Natarajan and Kristian Kersting, "Symbolic Evaluation for Deep and Tractable Multivariate Poisson Distributions"
  • Shalini Ghosh, Patrick Lincoln, Ashish Tiwari and Xiaojin Zhu, "Trusted Machine Learning: Model Repair and Data Repair for Probabilistic Models"

14:45 - 15:25: Keynote 4: Rina Dechter, UC Irvine, "Probabilistic Reasoning Meets Heuristic Search on Max and Sum queries"


15:25 - 17:00: Coffee and Poster Session (all workshop papers)


Organizers

Vaishak Belle, University of Edinburgh 
(contact: belle.vaishak[at]gmail.com)

Rodrigo de Salvo Braz, SRI International 
(contact: braz[at]ai.sri.com)

Kristian Kersting, TU Dortmund 
(contact: kristian.kersting[at]cs.tu-dortmund.de)

Martin Mladenov, TU Dortmund
(contact: martin.mladenov[at]cs.tu-dortmund.de)

Scott Sanner, University of Toronto 
(contact: ssanner[at]mie.utoronto.ca)

Program Committee


Stefano Ermon, Stanford University

Vibhav Gogate, University of Texas at Dallas

Roni Khardon, Tufts University

Radu Marinescu, IBM Research

Wannes Meert, KU Leuven

Quoc-Sang Phan, Carnegie Mellon University

Important Dates

  • Submission deadline: November 15, 2016
  • Notification date: November 18, 2016
  • Camera-ready to AAAI: December 15, 2016
  • Workshop date: February 5, 2017 (one day)

Call for Submissions (deadline has passed)

Topics include (but are not limited to)
  • symbolic methods for inference (e.g., SMT)
  • symbolic approaches for handling both discrete and continuous probability spaces
  • symbolic planning and scheduling approaches
  • symbolic approaches to bettٰer represent and solve optimization problems
  • symbolic and algebraic methods in machine learning
Because purely logical inference is well-covered at AI, we would like to de-emphasize such submissions unless they cover probabilistic, mixed discrete/continuous, arithmetic, optimization, or other novel uses / expressive extensions of logical inference.  Please contact the organizers if unsure about your potential submission's relevance for this workshop.

Submission Procedure (deadline has passed)

We welcome previously unsubmitted work, papers submitted to the main AAAI conference, and papers reporting research already published provided they align well with the workshop topic.  

Three types of submissions are solicited:
  1. full-length papers (up to 6 pages + 1 page for references in AAAI format)
  2. challenge or position papers (2 pages + 1 page for references in AAAI format)
  3. already published papers (1 page: an abstract in AAAI format with a link to the full paper)
Paper submissions should be made through the workshop EasyChair web site: https://easychair.org/conferences/?conf=syminfopt17