Hyperspace geography: visualizing fitness landscapes beyond 4D

Area of Science:

• Computational Science
• Data Visualization
• Human-Computer Interaction

Background:

• Human perception excels at understanding 4D space-time but struggles with higher dimensions.
• Developing intuition for complex spatial structures necessitates leveraging perceptual and cognitive abilities.
• Active navigation and exploration are key to understanding intricate data landscapes.

Purpose of the Study:

• To present a technique for visualizing surfaces in moderate-dimensional binary spaces.
• To introduce the Hyperspace Graph Paper (HSGP) tool for interactive exploration.
• To demonstrate HSGP's utility in evolutionary computation for analyzing fitness landscapes and search algorithms.

Main Methods:

• Recursive unfolding of surfaces onto a 2D hypergraph.
• Interactive visualization using the freely available Web-based tool, HSGP.
• Application to continuous functions over Boolean variables in 4 to 16 dimensions.

Main Results:

• HSGP facilitates active user exploration of fitness landscapes.
• The tool allows mapping neighborhood structures and identifying global properties like basins of attraction.
• Effective for visualizing recursive and repetitive landscapes, aiding intuition development.

Conclusions:

• The hypergraph unfolding technique effectively visualizes complex high-dimensional spaces.
• HSGP empowers users to develop intuitive understanding through active navigation and pattern detection.
• This approach is particularly powerful for exploring evolutionary computation landscapes.