AAAI Fall Symposium on Flexible Computation

Home

Flexible Computation in Intelligent Systems: Results, Issues, and Opportunities

Flexible computation refers to procedures that allow a graceful trade-off to be made between the quality of results and allocations of costly resources, such as time, memory, or information. Systems employing flexible computation gain the ability to adapt the quality of their response to dynamic changes in requirements for precision, and to uncertainty or variation in the cost of computational commodities. Recent examples of flexible computation techniques include memory-bounded search, anytime algorithms, approximate query processing, and a variety of imprecise computation techniques. Flexible computation has been applied to combinatorial optimization, planning, probabilistic inference, decision making, and theorem proving. Our goal was to explore results, critical problems, and opportunities via invited talks, presentation papers, and panel discussions.

Fall and Spring Symposia are sponsored by the American Association for Artificial Intelligence (AAAI).

Flexcomp Program Committee

Program

The program for the Fall Symposium on Flexible Computation included invited talks, short research talks (15 minutes including questions), panels, and working groups.

Saturday, November 9th

Time Session Speaker
9:00–10:30

Session I

9:00–9:05

Welcome and Introduction

Eric Horvitz and Shlomo Zilberstein
9:05–9:50

Invited Talk: Resource Tradeoffs in Planning, Reasoning, and Search

Bart Selman
Directions in Flexible Computation

9:50–10:05

Issues in Design-to-time Real-Time Scheduling Alan Garvey and Victor Lesser
10:05–10:20

Models of Continual Computation in Automated Problem Solving

Eric Horvitz
10:20–10:30
Discussants and Open Discussion
10:30–11:00

Break

11:00–12:45

Session II
Flexibility in Planning and Search

11:00–11:15

State-Space Reduction for Flexible and Approximate Computation Weixiong Zhang

11:15–11:30

Anytime heuristic search: Preliminary Report Eric Hansen and Shlomo Zilberstein

11:30–11:45

Composing and Monitoring Non-Deterministic Design-to-Time Methods François Charpillet and Anne Boyer

11:45–12:00

The Tradeoff Between Speed and Optimality in Hierarchical Search Robert Holte, Maria Perez, Robert Zimmer, and Alan MacDonald

12:00–12:15

A Decision-Theoretic Approach to Designing High Quality Anytime Algorithms Guiyou Qiu

12:15–12:30

Discussants and Open Discussion

12:30–2:00

Lunch

2:00–3:30

Session III

2:00–2:45

Panel Discussion: Frontiers in Utility-Directed Search
Exploring Flexible Computation in Applications I

2:45–3:00

Flexible Systems for Digital Signal Processing Joseph Winograd, Jeffrey Ludwig, S. Hamid Nawab, Alan Oppenheim, and Anantha Chandrakasan

3:00–3:15

Gathering Information Based on its Value Joshua Grass and Shlomo Zilberstein

3:15–3:30

Operating System Support for Imprecise Computation David Hull, Wu Feng, and Jane W. S. Liu

3:30–4:00

Break

4:00–5:30

Session IV

4:00–4:30

Panel: What unifying principles of flexible computation provide value across applications?
Flexible Computation in Decision Making

4:30–4:45

Some Experiments with Real-Time Decision Algorithms Bruce D’Ambrosio and Scott Burgess

4:45–5:00

Flexible Policy Construction by Information Refinement Michael Horsch and David Poole

5:00–5:15

Model Selection under Time Constraints Geoffrey Rutledge

5:15–5:30

Evidence-directed belief network simplification Daphne Koller

 

Sunday, November 10th

Time Session Speaker
9:00–10:30

Session V

9:00–9:05

Overview of Second Day

Eric Horvitz and Shlomo Zilberstein
9:05–9:50

Invited Talk: Approximate Computation

Victor Lesser
Flexible Computation and Learning Strategies

9:50–10:05

An Anytime Solution to Inductively Learn Decision Trees Richard Jensen and Jeffrey Schlimmer
10:05–10:20

Supporting Flexibility: A Case-Based Reasoning Approach

Igor Jurisica
10:20–10:30
Discussants and Open Discussion
10:30–11:00

Break

11:00–12:30

Session VI
Exploring Flexible Computation in Applications II

11:00–11:15

Performing Concurrent Robot Path Execution and Computation in Real-Time John Zelek

11:15–11:30

Flexible Computation for 3D Graphics Under Varying Resources Eric Horvitz and Jed Lengyel

12:00–12:15

Handling Tradeoffs between Precision and Robustness for Visual Tracking Kentaro Toyama

12:15–12:30

An Admission Control Paradigm for Value-cognizant Real-Time Databases Azer Bestavros and Sue Nagy

12:30–2:00

Lunch

2:00–3:30

Session VII

2:00–2:45

Panel: What have we learned from applications of flexible computation?
Flexible Computation and Cognition

2:45–3:00

Flexible Computations Underlying Cognitive Plasticity Itiel Dror and Michael Young
3:00–3:15 Problem-Solving under Insufficient Resources Pei Wang

3:15–3:30

Discussants and Open Discussion

3:30–4:00

Break

4:00–5:30

Session VIII
Division into Focus Groups

Monday, November 11th

Time Session Speaker
9:00–10:30

AAAI Joint Session: Fall Symposium Invited Speaker

10:30–11:00

Break

11:00–12:30

Session IX

11:00–12:00 Focus Group Reports and Discussion

12:00–12:30

Panel: What are the key open research challenges in flexible computation?

Symposium Wrap-up