SMI'11 SMI'11


Invited Speaker: Chen Greif

University of British Columbia

Iterative Solvers for Linear Systems: A Computational Bottleneck and How to Possibly Deal with it...

Computer graphics models often raise the need to rapidly solve a linear system or a sequence of linear systems of equations. Such systems are frequently very large and sparse, and solving them numerically may form a major computational bottleneck. Direct solvers based on Gaussian elimination are powerful in general, but may not always give us the efficiency we are seeking. In particular, they cannot fully exploit the structure and sparsity of the problem, and do not allow us to seamlessly incorporate a good initial guess. In this talk I will provide an overview of modern iterative solvers. Such solvers are suitable for dealing with sparsity and large scale, yet making them effective is a challenging task. We will focus in particular on structured problems with constraints, which give rise to saddle-point systems. A crucial aspect in the successful implementation of these solvers is the use of an effective preconditioning technique, which aims to accelerate convergence by changing the matrix associated with the linear system. A few illustrative examples of applications will be given.

Invited Speaker: Hao (Richard) Zhang

Simon Fraser Universtity

Towards High-Level Geometry Processing

Low-level geometry processing focuses on the computation and utilization of local measurements of shape geometry. As the field of computer graphics moves forward, we are naturally progressing from merely treating shapes at an elementary and purely geometric level towards analyzing and manipulating them at a more structural and even semantic level; I refer to the latter as high-level geometry processing. In this talk, I will cover three latest developments towards high-level geometry processing. First, I will show how symmetry, a purely geometric notion, enables shape analysis at a more global and structural level to produce more meaningful shape segmentation. Second, I will show how shape semantics can be incorporated to facilitate shape correspondence amid large geometric discrepancies between corresponding parts. At last, I will show benefits of utilizing a set of shapes for shape analysis and synthesis. What is common about all these approaches is the utilization of prior knowledge, albeit of different kinds and in different ways.

First Day

Opening Session 9:00-9:30

Key 9:30-10:30 Chen Greif (University of British Columbia)

Session1 11:00-12:15

Session2 13:45-15:30

Session3 16:00-17:15

Second Day

Key 9:30-10:30 Hao (Richard) Zhang (Simon Fraser Universtity)

Session1 11:00-12:15

Session2 13:45-15:30

Session3 16:00-17:40