Researching recurrent neural network structures in the brain


Two researchers from the University of Wyoming decided to pluck each other’s brains, so to speak. In particular, they examined the importance of the frontal cortex, the part of the brain used for decision making, expressive language, and voluntary movement.

And the two scientists learned that a recurrent neural network structure (RNN) is responsible for these functions.

“This RNN receives inputs from emotional regions of the brain and sends outputs to the motor cortex, the part of the brain that is responsible for voluntary movement,” says Qian-Quan Sun, a UW professor of zoology and physiology. “In the field of artificial intelligence, computer scientists have designed various artificial neural networks, including RNNs, that effectively solve problems such as language translation and object recognition by simulating the neural network in the mammalian brain.

“This paper provides a basic structure of neural networks in the mammalian brain. That framework will guide us as we investigate behavioral strategies, ”Sun continues. “After we have received more details, we can translate it into an artificial neural network and use it to solve real problems.”

Sun, director of UW’s Wyoming Sensory Biology Center of Biomedical Research Excellence, is the lead author of a paper titled “A Long-Range Recurrent Neuronal Network Linking the Emotion Regions with Somatic Motor Cortex,” published Tuesday in Cell Reports . The open access journal publishes peer-reviewed articles across the spectrum of life sciences that report on new biological findings.

The paper’s first author is Yihan Wang, a Ph.D. Student in the Doctoral Neuroscience Program of the UW from Beijing, China. The research was funded by grants from the National Institutes of Health.

Artificial RNNs are important deep learning algorithms that are often used for ordinal or temporal lobe problems, such as: B. Language translation, natural language processing, speech recognition and captioning, says Sun. An RNN recognizes sequential features in data and uses patterns to predict the next likely scenario. RNNs are built into popular applications like Siri, Google Voice Search, and Google Translate.

“The biggest surprise is that RNNs not only exist in our brains, but that they are constructed with much finer functions and yet are highly efficient at processing sequential inputs,” says Sun. “In general, cortical neurons are spatially reciprocal and mix with one another. However, Wang’s data not only indicated that the RNN exists in the most important part of the brain – the frontal cortex – but that this network is also less complex than we thought and largely unidirectional. This comes as a great surprise to us as it tells us that this network may be responsible for unique functions compared to others. “

Sun and Wang analyzed the brains of mice for laboratory research. Various genetically modified mouse strains gave the two the ability to mark certain types of neurons with fluorescent proteins that follow the connections of the brain – and to monitor the activities of certain neurons with intrinsically fluorescent markers.

The research has many real world implications, according to Sun.

“For one thing, now that we know about this important building block, the work will help further decipher how our brains make decisions,” he says. “Second, it will help uncover other similar RNAs in other parts of the brain. It will help researchers use computer simulations to predict how our brains encode short-term memory and how it can be used. And third, especially for this study, it will help us understand how emotions like fear and anxiety regulate our movements. “

Both the content and the research approach of Sun and Wang should have a very broad interest among artificial intelligence researchers, biologists, computer modelers, and neuroscientists, Sun says.

“The exact connection map can also help us understand the cause of the neurological and psychiatric disorders that have problems regulating emotions or voluntary movements,” says Sun. “Before this result can find broader applications, however, many details have to be found out – such as how the local inhibitory network has refined the RNN and how various components underlie certain emotional states.”

Wang’s goal is to work out these details in his dissertation, says Sun.


  1. Yihan Wang, Qian-Quan Sun. An extensive, recurrent neural network that connects the regions of emotion with the somatic motor cortex. Cell Reports, 2021; 36 (12): 109733 DOI: 10.1016 / j.celrep.2021.109733

/ Public release. This material is from the original organization (s) and may be of a sporadic nature that has been edited for clarity, style and length. The views and opinions expressed are those of the author (s). View in full here.


Leave A Reply