Matthew Chalmers
Research Scientist
Ubilab, Union Bank of Switzerland
Matthew.Chalmers@ubs.com
Information that is high-dimensional or otherwise complex in structure is increasingly common. At Ubilab we have been working on visualization techniques for complex information and applying them to textual documents, time series data, financial trade records and Excel spreadsheets. This talk will summarize our work on two complementary areas: firstly algorithms for finding 'map positions' in a low dimensional space (usually 2) which represent the similarities and patterns in a body of high dimensional data, and secondly the graphical representation and interaction with such a map. In the first area we have focused on fast layout algorithms based on spring models and simulated annealing. In the second, issues such as maintaining legibility and showing traces of past activity have been foremost. Lastly we discuss the modifications and additions which have been involved in doing this research collaboratively with a trading division of the bank.