How does the brain efficiently use resources?


A collaboration between SISSA and the University of Pennsylvania indicates the existence of an efficient sensory coding process in rats, suggesting a general principle for optimal use of computing resources. The study, published in eLife, opens the way to the understanding of the neural mechanisms underlying this efficiency and to the development of artificial intelligence systems based on similar principles.

The brains of human beings at rest are said to use one-fifth of the energy produced by their body, or about 20 W of power. In fact, the cost would be much higher if our brains were not equipped with an efficient coding mechanism that allows us to represent only really useful information in the vast continuous flow of sensory stimuli. A new study by SISSA and the University of Pennsylvania, published in eLife, shows the existence of similar efficient coding processes for visual stimuli in rodents. The results support the important theory of efficient coding of sensory perception and pave the way for new experimental approaches to understand the underlying neural mechanisms and the development of new training protocols for machine vision systems.

In the early 1960s, British scientist Horace Barlow proposed the hypothesis of efficient coding. “According to this theory, our brain can efficiently represent sensory information using a neural code that minimizes the number of impulses, and therefore the energy, required to encode and transmit information,” explains Davide Zoccolan, director from the visual neuroscience laboratory at La Scuola. Internazionale Superiore di Studi Avanzati – SISSA. “This especially occurs in the visual system, due to the decreasing number of neurons in the deeper areas of the cortex which reduces the representational capacity.”

According to information theory, which underlies the efficient coding hypothesis, an efficient sensory system should preferably allocate computational resources to represent statistical characteristics of the environment that are more informative about its state. In the case of the visual system, this means encoding the most informative features of the natural images around us.

Vijay Balasubramanian, a computational neuroscientist at the University of Pennsylvania, has been working on this topic for the past decade: “We have analyzed thousands of images of natural landscapes turning them into binary images, made up of black and white pixels, and breaking them down. in different textures defined by specific statistics ”, explains the researcher. “We have noticed that different types of textures have different variability in nature and that human subjects are better able to recognize which ones vary the most. It is as if our brains are allocating resources where they are needed most.

Until now, there was no evidence that such effective perceptions of visual textures occur in other species. In a new study published in eLife, the Zoccolan and Balasubramanian teams established that the effect occurs in rodents.

Riccardo Caramellino, first author of the study, along with Andrea Buccellato and Anna Carboncino, trained the animals to discriminate binary images consisting of random black and white pixels from textures created according to specific probabilistic criteria, as was done previously with human subjects. They then analyzed the results using a mathematical model of an “ideal observer” developed by Eugenio Piasini, co-first author of the article. Scientists have observed that rodents, like humans, are the most sensitive to textures which vary the most in nature.

“We have found, in rodents, a model of perceptual sensitivity for visual textures that is consistent with efficient coding and is the same as that previously observed in humans, despite the phylogenetic distance between these species. This result suggests that efficient texture coding may be a universal principle in vision, ”comments Zoccolan. “The visual system seems to adapt to the surrounding environment by a kind of passive exposure, specializing in the recognition of more informative signals, thus allowing a considerable saving of computational resources and energy. Our study paves the way for new experimental approaches to study the neural mechanisms behind this fundamental process. It also suggests new ways to train artificial vision systems, based on the same principle. “

Reference: Caramellino R, Piasini E, Buccellato A, Carboncino A, Balasubramanian V, Zoccolan D. The sensitivity of rats to multipoint statistics is predicted by efficient coding of natural scenes. eLife. 2021; 10: e72081. doi: 10.7554 / eLife.72081

This article was republished from the following materials. Note: Material may have been edited for length and content. For more information, please contact the cited source.


Comments are closed.