Multiple Representations in geospatial databases, the brain's spatial cells, and deep learning algorithms.

Area of Science:

• Geographic Information Science (GIScience)
• Spatial Cognition
• Deep Learning

Background:

• Pioneering research by Buttenfield (1988) introduced multiple representations in GIScience.
• Multiple representations address challenges in abstracting geographic complexity for spatial databases and cartography.
• These issues include ontological and implementational complications within geographic information systems (GIS).

Purpose of the Study:

• To review multiple representations in spatial databases, spatial cognition, and deep learning, expanding on Buttenfield's foundational work.
• To explore how multiple representations, initially seen as a hindrance, can encode and decipher geographic complexity.
• To synthesize literature on cognitive and feature representations to understand their role in learning geography.

Main Methods:

• Literature synthesis across GIScience, spatial cognition (hippocampal formation), and deep learning.
• Cross-referencing concepts of multiple representations in GIScience, brain spatial cells, and machine learning algorithms.
• Acknowledging Buttenfield's contributions to multiple representations in GIScience.

Main Results:

• Multiple representations provide novel perspectives for encoding and deciphering geographic complexity.
• Cognitive representations of space exist in the brain's hippocampal formation.
• Deep learning utilizes feature representations for spatial data.

Conclusions:

• Multiple representations are beneficial, not detrimental, in GIS.
• There are parallels between cognitive spatial representations in humans and feature representations in deep learning.
• Multiple representations facilitate the learning of geography for both humans and machines.