I have seen the use of orthogonal coding to separate and protect information in the brain before.For example, when the monkey is preparing to exercise, the neural activity in the motor cortex represents the underlying movement, but This is to avoid interference The actual commands are transmitted to the muscles through signals.
Nevertheless, it is generally unclear how neural activity is transformed in this way. Bushman and Libby wanted to answer this question to understand what they observed in the auditory cortex of mice. Libby said: “When I first started working in the laboratory, it was hard for me to imagine that such things would happen with neural stimulation activity.” She wanted to “turn on the work done by neural networks to create this orthogonality.” Black box”.
Through experimental screening of various possibilities, they ruled out the possibility that different subsets of auditory cortex neurons independently process sensory and memory representations. Instead, they show that they participate in the same general neuron population, and the activity of neurons can be neatly divided into two categories. Some “stable” behaviors during sensory and memory representations, while other “switching” neurons flip their response patterns each time they are used.
To the researchers’ surprise, this combination of stabilizing and switching neurons is sufficient to rotate sensory information and convert it into memory. “That’s all the magic,” Bushman said.
In fact, he and Libby (Libby) used computational modeling methods to prove that this mechanism is the most effective way to construct orthogonal representations of sensation and memory: it requires fewer neurons and more than other methods. Less energy.
The discovery of Buschman and Libby has promoted the emerging trend of neuroscience: even in the lower sensory area, the dynamic coding of the neuron group is more abundant than previously thought. “These parts of the cortex in the lower part of the food chain also have very interesting dynamics, and maybe we haven’t really appreciated it until now,” he said. Miguel MaravallHe is a neuroscientist at the University of Sussex, and he is not involved in this new research.
This work may help reconcile two aspects of the ongoing debate about whether short-term memory is maintained through constant, lasting representations or through dynamic neural codes that change over time. Bushman said: “Our results show that they are basically right, rather than unilaterally decline,” stable neurons achieve the former, while switching neurons. The combination of processes is useful because “it actually helps prevent interference and perform orthogonal rotation.”
The research of Buschman and Libby may be related to backgrounds other than sensory representations. They and other researchers hope to find the mechanism of this orthogonal rotation in other processes: how the brain tracks multiple ideas or targets at the same time; how to complete tasks while processing interference; how it represents internal states; how to control cognition, including Attention process.
“I’m really excited,” Bushman said. Looking at the work of other researchers, “I just remember seeing that there is a stable neuron and a switching neuron! You can see them everywhere now.”
Libby is interested in the implications of their results for artificial intelligence research, especially the design of useful architectures for AI networks that must perform multiple tasks. She said: “I want to see if people pre-allocate neurons in the neural network to make it stable and switching characteristics, not just random characteristics, which helps their network to some extent.”
All in all, “it will be very important to figure out the consequences of this information encoding, and very interesting,” Maravall said.
ability Reprinted with permission below Quanta Magazine, Edit independent publications Simmons Foundation Its mission is to enhance the public’s understanding of science by covering research developments and trends in mathematics, physics and life sciences.
More exciting wired stories