Papers about visual number sense in deep neural nets
Humans and animals have a “number sense,” an innate capability to intuitively assess the number of visual items in a set, its numerosity. This capability implies that mechanisms to extract numerosity indwell the brain’s visual system, which is primarily concerned with visual object recognition.
Here, we show that network units tuned to abstract numerosity, and therefore reminiscent of real number neurons, spontaneously emerge in a biologically inspired deep neural network that was merely trained on visual object recognition.
These numerosity-tuned units underlay the network’s number discrimination performance that showed all the characteristics of human and animal number discriminations as predicted by the Weber-Fechner law. These findings explain the spontaneous emergence of the number sense based on mechanisms inherent to the visual system.
Visual number sense in untrained deep neural networks (Kim et al., 2021)
Number sense, the ability to estimate numerosity, is observed in naïve animals, but how this cognitive function emerges in the brain remains unclear.
Here, using an artificial deep neural network that models the ventral visual stream of the brain, we show that number-selective neurons can arise spontaneously, even in the complete absence of learning. We also show that the responses of these neurons can induce the abstract number sense, the ability to discriminate numerosity independent of low-level visual cues. We found number tuning in a randomly initialized network originating from a combination of monotonically decreasing and increasing neuronal activities, which emerges spontaneously from the statistical properties of bottom-up projections.
We confirmed that the responses of these number-selective neurons show the single- and multineuron characteristics observed in the brain and enable the network to perform number comparison tasks. These findings provide insight into the origin of innate cognitive functions.