Canonical Correlation Analysis: Use of Composite Heliographs for Representing Multiple Patterns
- Cite this paper as:
- Degani A., Shafto M., Olson L. (2006) Canonical Correlation Analysis: Use of Composite Heliographs for Representing Multiple Patterns. In: Barker-Plummer D., Cox R., Swoboda N. (eds) Diagrammatic Representation and Inference. Diagrams 2006. Lecture Notes in Computer Science, vol 4045. Springer, Berlin, Heidelberg
In a study of crew interaction with the automatic flight control system of the Boeing 757/767 aircraft, we observed 60 flights and recorded every change in the aircraft control modes, as well as every observable change in the operational environment. To quantify the relationships between the state of the operating environment and pilots’ actions and responses, we used canonical correlation because of its unique suitability for finding multiple patterns in large datasets. Traditionally, the results of canonical correlation analysis are presented by means of numerical tables, which are not conducive to recognizing multidimensional patterns in the data. We created a sun-ray-like diagram (which we call a heliograph) to present the multiple patterns that exist in the data by employing Alexander’s theory of centers. The theory describes 15 heuristic properties that help create wholeness in a design, and can be extended to the problem of information abstraction and integration as well as packing of large amounts of data for visualization.
Unable to display preview. Download preview PDF.