(ORDO NEWS) — A study published in the scientific journal Nature Neuroscience reports that the brain applies data compression when it makes a decision.
Senior author Joe Paton, director of the Champalimaud Neuroscience Research Program, says that the idea that the brain maximizes performance while minimizing cost through the use of data compression is pervasive in sensory processing research. However, it has not yet been explored in the field of cognitive functions.
“Using a combination of experimental and computational methods, we have demonstrated that the same principle applies to a much broader range of functions than previously thought,” he noted.
In their experiments, the researchers used the timing paradigm. In each trial, the mice had to determine whether the two tones were separated by more or less than 1.5 seconds. At the same time, the scientists recorded the activity of dopamine neurons in the animals’ brains during the task.
“It is well known that dopamine neurons play a key role in learning the value of actions,” the experts explained.
Therefore, if an animal misjudged the length of an interval in a given trial, the activity of these neurons resulted in a “prediction error” that should help improve performance in future trials.
Several computational reinforcement learning models were built to test which one best represented neuronal activity and animal behavior. The models shared some common principles, but differed in how they represented information that might be important to completing a task.
The team found that only models with a condensed view of the problem could explain the results.
“The brain seems to weed out all irrelevant information. Interestingly, it also gets rid of some relevant information, but not so much that it greatly affects the total amount of reward that the animal receives.
The animal clearly knows how to succeed in this game,” experts say.
According to the authors, this discovery has important implications for both neuroscience and artificial intelligence.
“Although the brain has clearly evolved to process information efficiently, AI algorithms often solve problems by brute force: using a lot of data and a lot of parameters.
Our work offers a set of principles that can guide future research into how internal representations of the world can be maintained,” the scientists stressed.
Contact us: [email protected]