2024-07-04 17:00:00
Memories serve two purposes: they allow us to remember specific events from the past, but they also allow us to extract general ideas from them. For example, we can remember specific details from a recent visit to a friend, but we can also draw his appearance and appearance. A region of the brain known as the hippocampus is responsible for both abilities, but it is unclear how it does this.
Interestingly, the hippocampus receives two inputs: one in which experiences come directly from the brain region above it; and another in which they make a return through another brain region, where they are processed.
To test this hypothesis, scientists at the RIKEN Institute used computational physics techniques to create a mathematical model of the neural circuits of the hippocampus. Then they tested the model using pictures of three items of clothing: sneakers, pants and coats. They discovered that the first method saved the details of each image, while the second generated general conceptual images of each type of clothing.
At this stage, the model provides a possible description of what may be happening in the hippocampus, but more evidence is needed to support it. To confirm this, the researchers tested their model with experimental recordings of hippocampal neurons from mice. They note that the signature generated by the model is also present in the recordings. Finally, they apply the same approach to machine learning, so neural networks represent information in both symbolic and conceptual form. This results in better performance in distinguishing between similar images and identifying common features.
This work was published in the journal Nature Communication and Physical Review E and shows how the understanding of physical learning can inspire better algorithms for artificial intelligence.
#hippocampus #pathways #integrate #memory