-
Categories
-
Pharmaceutical Intermediates
-
Active Pharmaceutical Ingredients
-
Food Additives
- Industrial Coatings
- Agrochemicals
- Dyes and Pigments
- Surfactant
- Flavors and Fragrances
- Chemical Reagents
- Catalyst and Auxiliary
- Natural Products
- Inorganic Chemistry
-
Organic Chemistry
-
Biochemical Engineering
- Analytical Chemistry
-
Cosmetic Ingredient
- Water Treatment Chemical
-
Pharmaceutical Intermediates
Promotion
ECHEMI Mall
Wholesale
Weekly Price
Exhibition
News
-
Trade Service
The iNature human brain processes sequence information all the time, whether it is language communication, action implementation or episodic memory, all of which essentially involve the representation of sequential information
.
On the other hand, the execution of a sequence takes a certain amount of time, and the brain needs to memorize the entire sequence before applying the timing information
.
For example, we need to remember a series of directions given by the guide when asking for directions, and remember a series of movement patterns demonstrated by the teacher when learning new dance moves
.
In these cases, not only the individual items need to be remembered, but the order between them must not be confused
.
Cognitive psychologists began to think about the representation of sequence information as early as the early 19th century.
The coding of sequence information is also considered to be the premise of the syntactic structure of human language.
model
.
However, little is known about the neural coding mechanisms of the brain with memory of temporal information
.
On February 10, 2022, the Center for Excellence in Brain Science and Intelligent Technology of the Chinese Academy of Sciences (Institute of Neuroscience), the Wang Liping Research Group of the Key Laboratory of Primate Neurobiology, Chinese Academy of Sciences, and the Deputy Min Bin of the Shanghai Center for Brain Science and Brain-like Research The researcher and Tang Shiming's group from the School of Life Sciences of Peking University jointly published a research paper titled "Geometry of sequence working memory in macaque prefrontal cortex" in Science Online.
Two-photon calcium imaging technology records neuronal activity in the prefrontal cortex of the macaque brain
.
The study found that neurons represented each spatial position in the sequence in a population-encoded form, and found similar loop geometries in these representations
.
The study overturns key assumptions of the classical sequential working memory model, providing new insights into the difficult problem of how neural networks perform symbolic representations
.
The human brain processes sequential information all the time, whether it is language communication, action implementation or episodic memory, all of which are essentially related to the representation of sequential information
.
On the other hand, the execution of a sequence takes a certain amount of time, and the brain needs to memorize the entire sequence before applying the timing information
.
For example, we need to remember a series of directions given by the guide when asking for directions, and remember a series of movement patterns demonstrated by the teacher when learning new dance moves
.
In these cases, not only the individual items need to be remembered, but the order between them must not be confused
.
Cognitive psychologists began to think about the representation of sequence information as early as the early 19th century.
The coding of sequence information is also considered to be the premise of the syntactic structure of human language.
model
.
However, little is known about the neural coding mechanisms of the brain with memory of temporal information
.
Macaque is the model animal closest to humans in evolution.
Its cognitive ability, brain structure and function are closer to humans than other model animals.
It is the best model for studying complex advanced cognitive functions such as time series
.
Therefore, to explore the problem of temporal memory encoding, the researchers trained macaques to memorize spatial sequences consisting of multiple location points (Fig.
1)
.
During the task, three different dots flashed on the screen in front of the macaques in sequence, and the macaques needed to report the dots in the order presented a few seconds later
.
During the memory retention period of a few seconds before reporting, the spatial sequence of information is temporarily stored in the brain in the form of working memory
.
To record the activity of brain neuronal populations during tasks in macaques, the researchers imaged two-photon calcium signals in the lateral prefrontal cortex, the home of working memory
.
Calcium signals can reflect the firing activity of neurons, and the key to the representation of sequence information lies in the activity patterns of neuronal populations during memory periods
.
Figure 1: How does the brain of a macaque monkey spatial sequence memory task simultaneously represent multiple pieces of information in a sequence during the memory period? The researchers speculate that the macaques also have a "screen" in their brains on which the macaques can memorize points that appear
.
But if three points are displayed on this screen at the same time during the memory retention period, how should the order of each point be reflected? Could there be three different screens in the brain of a macaque at the same time? In this way, each screen only needs to record the information of one point, and the screens will not interfere with each other
.
The researchers analyzed the high-dimensional data obtained by calcium imaging and found that the two-dimensional subspace (subspace) corresponding to the information of each order can be found in the high-dimensional vector space, that is, to find its corresponding "screen" (Figure 2)
.
In each subspace, the spatial positions corresponding to different points are consistent with the ring structure of real visual stimuli
.
Moreover, the subspaces corresponding to different orders are nearly orthogonal to each other, indicating that the brain does use three different screens to represent sequence information
.
In order to further explore whether the brain always uses the same "screens" to memorize different types of spatial sequences, the researchers did a decoding analysis of the data, that is, using machine learning methods to train a linear classifier to distinguish spatial information in different orders
.
For example, training the decoder with the neuron group activity when the macaque responds correctly can achieve better decoding results in partially correct sequences
.
These results suggest that the "screen" used for coding order is stable and general
.
Figure 2: Representation of sequence memory in neural high-dimensional vector space The researchers also found that similar ring structures are shared between subspaces of different orders, except that the size of the ring radius decreases as the order increases
.
A possible explanation is that less attention resources are allocated to the information in the later order, resulting in a smaller corresponding loop and a lower degree of discrimination
.
This structure also corresponds to the behavioral performance of sequential memory.
For example, if we remember more content in our daily life, the later information is more prone to errors
.
This finding can also be summarized as the temporally modulated nature of the spatial information encoding geometry at the population level
.
Interestingly, this property does not fully apply at the single neuron level, and the enhanced modulation of single neuron activity is the key assumption of the classical sequential working memory model, suggesting that the encoding of sequential memory should pay more attention to the group neuron property
.
Figure 3: Researcher Wang Liping, Associate Researcher Min Bin, Research Assistant Hu Peixin, and Postdoctoral Fellow Xie Yang discussing the topic.
This research is the first time to explain the calculation and coding principle of sequential working memory at the population neuron level, and also to explain how neural networks perform symbolic representation.
A puzzle offers new ideas
.
In the 1980s, some researchers in the field of artificial intelligence proposed the concept of tensor product to realize the representation of symbolic structure in neural networks, but how it emerges naturally at the neural network level has not been well resolved
.
The neural representation of sequential working memory just corresponds to embedding the symbolic representation into a high-dimensional vector space from a subspace of corresponding order, and supports the linear reading of symbolic structural information by downstream neural networks
.
Xie Yang, postdoctoral fellow of Wang Liping's research group at the Center for Excellence in Brain Science and Intelligent Technology, Chinese Academy of Sciences, and research assistant Hu Peizhen are the co-first authors of the paper.
Research assistant Li Junru, doctoral student Chen Jingwen, and Peking University graduate student Song Weibin made contributions at different stages of the project
.
Professor Stanislas Dehaene of the French National Institute of Health and Medical Research, Professor Wang Xiaojing of New York University and researcher Yang Tianming of the Center of Excellence for Brain Intelligence, Chinese Academy of Sciences participated in this study
.
This work was funded by the Chinese Academy of Sciences, the Foundation Committee, the Ministry of Science and Technology, and the Shanghai Municipal Government
.
Note: The analysis refers to the introduction from the official website of the Institute of Neurology, Chinese Academy of Sciences
.
Analysis link: http:// Reference message: https://
.
On the other hand, the execution of a sequence takes a certain amount of time, and the brain needs to memorize the entire sequence before applying the timing information
.
For example, we need to remember a series of directions given by the guide when asking for directions, and remember a series of movement patterns demonstrated by the teacher when learning new dance moves
.
In these cases, not only the individual items need to be remembered, but the order between them must not be confused
.
Cognitive psychologists began to think about the representation of sequence information as early as the early 19th century.
The coding of sequence information is also considered to be the premise of the syntactic structure of human language.
model
.
However, little is known about the neural coding mechanisms of the brain with memory of temporal information
.
On February 10, 2022, the Center for Excellence in Brain Science and Intelligent Technology of the Chinese Academy of Sciences (Institute of Neuroscience), the Wang Liping Research Group of the Key Laboratory of Primate Neurobiology, Chinese Academy of Sciences, and the Deputy Min Bin of the Shanghai Center for Brain Science and Brain-like Research The researcher and Tang Shiming's group from the School of Life Sciences of Peking University jointly published a research paper titled "Geometry of sequence working memory in macaque prefrontal cortex" in Science Online.
Two-photon calcium imaging technology records neuronal activity in the prefrontal cortex of the macaque brain
.
The study found that neurons represented each spatial position in the sequence in a population-encoded form, and found similar loop geometries in these representations
.
The study overturns key assumptions of the classical sequential working memory model, providing new insights into the difficult problem of how neural networks perform symbolic representations
.
The human brain processes sequential information all the time, whether it is language communication, action implementation or episodic memory, all of which are essentially related to the representation of sequential information
.
On the other hand, the execution of a sequence takes a certain amount of time, and the brain needs to memorize the entire sequence before applying the timing information
.
For example, we need to remember a series of directions given by the guide when asking for directions, and remember a series of movement patterns demonstrated by the teacher when learning new dance moves
.
In these cases, not only the individual items need to be remembered, but the order between them must not be confused
.
Cognitive psychologists began to think about the representation of sequence information as early as the early 19th century.
The coding of sequence information is also considered to be the premise of the syntactic structure of human language.
model
.
However, little is known about the neural coding mechanisms of the brain with memory of temporal information
.
Macaque is the model animal closest to humans in evolution.
Its cognitive ability, brain structure and function are closer to humans than other model animals.
It is the best model for studying complex advanced cognitive functions such as time series
.
Therefore, to explore the problem of temporal memory encoding, the researchers trained macaques to memorize spatial sequences consisting of multiple location points (Fig.
1)
.
During the task, three different dots flashed on the screen in front of the macaques in sequence, and the macaques needed to report the dots in the order presented a few seconds later
.
During the memory retention period of a few seconds before reporting, the spatial sequence of information is temporarily stored in the brain in the form of working memory
.
To record the activity of brain neuronal populations during tasks in macaques, the researchers imaged two-photon calcium signals in the lateral prefrontal cortex, the home of working memory
.
Calcium signals can reflect the firing activity of neurons, and the key to the representation of sequence information lies in the activity patterns of neuronal populations during memory periods
.
Figure 1: How does the brain of a macaque monkey spatial sequence memory task simultaneously represent multiple pieces of information in a sequence during the memory period? The researchers speculate that the macaques also have a "screen" in their brains on which the macaques can memorize points that appear
.
But if three points are displayed on this screen at the same time during the memory retention period, how should the order of each point be reflected? Could there be three different screens in the brain of a macaque at the same time? In this way, each screen only needs to record the information of one point, and the screens will not interfere with each other
.
The researchers analyzed the high-dimensional data obtained by calcium imaging and found that the two-dimensional subspace (subspace) corresponding to the information of each order can be found in the high-dimensional vector space, that is, to find its corresponding "screen" (Figure 2)
.
In each subspace, the spatial positions corresponding to different points are consistent with the ring structure of real visual stimuli
.
Moreover, the subspaces corresponding to different orders are nearly orthogonal to each other, indicating that the brain does use three different screens to represent sequence information
.
In order to further explore whether the brain always uses the same "screens" to memorize different types of spatial sequences, the researchers did a decoding analysis of the data, that is, using machine learning methods to train a linear classifier to distinguish spatial information in different orders
.
For example, training the decoder with the neuron group activity when the macaque responds correctly can achieve better decoding results in partially correct sequences
.
These results suggest that the "screen" used for coding order is stable and general
.
Figure 2: Representation of sequence memory in neural high-dimensional vector space The researchers also found that similar ring structures are shared between subspaces of different orders, except that the size of the ring radius decreases as the order increases
.
A possible explanation is that less attention resources are allocated to the information in the later order, resulting in a smaller corresponding loop and a lower degree of discrimination
.
This structure also corresponds to the behavioral performance of sequential memory.
For example, if we remember more content in our daily life, the later information is more prone to errors
.
This finding can also be summarized as the temporally modulated nature of the spatial information encoding geometry at the population level
.
Interestingly, this property does not fully apply at the single neuron level, and the enhanced modulation of single neuron activity is the key assumption of the classical sequential working memory model, suggesting that the encoding of sequential memory should pay more attention to the group neuron property
.
Figure 3: Researcher Wang Liping, Associate Researcher Min Bin, Research Assistant Hu Peixin, and Postdoctoral Fellow Xie Yang discussing the topic.
This research is the first time to explain the calculation and coding principle of sequential working memory at the population neuron level, and also to explain how neural networks perform symbolic representation.
A puzzle offers new ideas
.
In the 1980s, some researchers in the field of artificial intelligence proposed the concept of tensor product to realize the representation of symbolic structure in neural networks, but how it emerges naturally at the neural network level has not been well resolved
.
The neural representation of sequential working memory just corresponds to embedding the symbolic representation into a high-dimensional vector space from a subspace of corresponding order, and supports the linear reading of symbolic structural information by downstream neural networks
.
Xie Yang, postdoctoral fellow of Wang Liping's research group at the Center for Excellence in Brain Science and Intelligent Technology, Chinese Academy of Sciences, and research assistant Hu Peizhen are the co-first authors of the paper.
Research assistant Li Junru, doctoral student Chen Jingwen, and Peking University graduate student Song Weibin made contributions at different stages of the project
.
Professor Stanislas Dehaene of the French National Institute of Health and Medical Research, Professor Wang Xiaojing of New York University and researcher Yang Tianming of the Center of Excellence for Brain Intelligence, Chinese Academy of Sciences participated in this study
.
This work was funded by the Chinese Academy of Sciences, the Foundation Committee, the Ministry of Science and Technology, and the Shanghai Municipal Government
.
Note: The analysis refers to the introduction from the official website of the Institute of Neurology, Chinese Academy of Sciences
.
Analysis link: http:// Reference message: https://
