Scale decomposition of a function

Large blocks correspond to fine scales while small blocks indicate coarse scales.

The datamining group at the Australian National University deals with massive datasets containing many independent variables. Searching for meaningful patterns requires efficient approximations to the original data that save memory but preserve important detail. The research is highly mathematical in nature. To make the fundamental concepts accessible to a semi-technical audience, the images at right were created by Darran Edmundson as a visual metaphor for an otherwise abstract topic.

LEFT: Compact representation of the original function

Discarding large blocks that do not add significantly to the original function yields a reasonable approximation using only a small fraction of the original information.

Copyright EDM STUDIO INC. 2006-2007