Computational foundations of the visual number sense

Area of Science:

• Cognitive Science
• Computational Neuroscience
• Developmental Psychology

Background:

• Numerosity perception, the ability to estimate quantities, is fundamental to cognition.
• Understanding the computational mechanisms and developmental trajectory of number sense is a key challenge in cognitive science.
• The role of inhibitory processes in refining numerical representations remains an active area of research.

Purpose of the Study:

• To propose an emergentist perspective on the computational basis of numerosity perception.
• To investigate the developmental aspects of number sense and the function of inhibition.
• To model the influence of continuous visual properties on numerosity perception using a deep neural network.

Main Methods:

• Development of a deep neural network model to simulate numerosity perception.
• Analysis of the model's performance in processing numerical and continuous visual information.
• Examination of how the model accounts for developmental changes and inhibitory effects.

Main Results:

• The deep neural network model provides an emergentist account of numerosity perception.
• Continuous visual properties were found to delineate the limits of, rather than invalidate, number sense computation.
• The model demonstrates how invariance in numerosity perception can be achieved.

Conclusions:

• The computational mechanism for numerosity perception emerges from the network's processing.
• The findings support an emergentist view, integrating continuous visual influences within a framework of number sense.
• Formalizing alternative theories in computational models is crucial for advancing the field.