Conference on Cognitive Computational Neuroscience

Across various domains in neuroscience, abundant evidence reveals individual neurons encoding specific, identifiable features of our external environment. This consistent response tuning of individual neurons suggests that brains may favor certain representational axes over others. But as yet, we do not understand whether and why brains use privileged bases for representing the natural world. Here, we develop a formal framework for investigating the extent to which a representational system has privileged axes. We formulate an axis-dependent alignment metric, which enables us to demonstrate that the axes of neural representations of sensory stimuli are in fact aligned across conspecifics. Parallel analyses of computational models of neurobiological systems reveals that Deep Convolutional Neural Networks also have privileged axes. Strikingly, we further observed that the representational axes of DCNNs trained on natural images and sounds are aligned with the privileged axes found in the visual and auditory cortices respectively. Additionally, our results suggest that the favored basis of brains and DCNNs results in higher lifetime sparseness and lower wiring costs downstream compared to an arbitrary basis, providing insight into the implications of this privileged axis. These findings offer a new computational framework for understanding the systematic neural tuning observed in distinct brain regions.