The memory process of the brain is doing data compression, and different graphics also have a common memory format

Alex from Concave Temple qubit | Public account How does the QbitAI brain remember the information it sees? Researchers have recently discovered that the brain captures visual information relevant to the current task and compresses what it sees; some irrelevant features are ignored
.

In other words, the memory process of the human brain is more like writing an outline on a blackboard than taking a picture with a camera
.

In addition, different visual information may be stored in nearly the same memory format is also the result of the researchers' recent experimental analysis
.

The paper, from New York University, has been published in Neuron, the top journal in neuroscience
.

Let's see what kind of research this is
.

Data collection of brain memory activity and visualization of the brain's ability to store and recall needed information, known as working memory, is essential for many higher-order cognitive processes in the human brain
.

Previous research has established that the human brain can re-encode visual information, for example, by re-encoding the numbers seen in the mind and storing them as sounds that read the numbers
.

Based on this conclusion, the research team conducted experiments to carry out new explorations
.

In the experiments, they set up two different images of visual stimuli: directional gratings and moving dots
.

In each test, the participants first saw an image, then spent 12 seconds recalling what they saw, and then used their memory to determine the direction in which the grating was tilted or the movement of a blob
.

Meanwhile, the researchers visualized the participants' brain activity using functional magnetic resonance imaging (fMRI)
.

The memory format of different visual information may be the same The researchers found that whether it is the tilt angle of the grating or the moving direction of the point, the same neural activity pattern is generated in the visual cortex and the parietal cortex
.

The visual cortex is the cerebral cortex that processes visual information, and the parietal cortex is the part of the brain that processes and stores memories
.

The specific data analysis results are as follows
.

The results showed that the visual cortex of the brain was able to distinguish between two different image stimuli, the tilt of the grating and the movement of the dots
.

At the same time, training the brain with one type of stimulus (such as a grating tilt angle) also allows it to successfully decode another type of stimulus (such as the direction of movement of a dot)
.

It is worth noting, however, that this "sharing" occurred only during the memory phase and was not observed during the time period when the stimulus was directly viewed
.

This also confirms a conjecture of the researchers: the image properties of raster tilt and point set movement are both different and similar, but in the process of working memory, the brain will only extract the features most relevant to the specific task, and will Two different stimuli are encoded into the same "storage format"
.

The way the human brain remembers is similar to "writing an outline".
The previous article proved that working memory is an abstract representation of image attributes in the brain
.

So, what are the potential properties of working memory representations? Does the brain store everything it sees? To dig deeper, the researchers used a sophisticated model that projected three-dimensional patterns of brain activity into a more informative two-dimensional representation of the visual space
.

This revealed the re-encoding format in which the brain stores and recalls information for the two visual stimuli in this experiment
.

Figure A is a schematic diagram of the spatial reconstruction analysis
.

The grating angle or point motion direction for each feature condition is projected into the field of view space
.

Figure B is the population reconstruction map.
According to Figure B, the filtering response and related fidelity values ​​are calculated to obtain Figure C and Figure
D.

Analyzing the data found that the participants actually didn't memorize gratings or complex moving point clouds at all; instead, they compressed the image into a line representing the angles they were asked to remember
.

About the authors Yuna Kwak, the first author of the paper, is currently pursuing a Ph.
D.
in psychology at New York University
.

Clayton E.
Curtis is the corresponding author of the paper, also from the NYU Department of Psychology
.

For now, researchers are just beginning to figure out some of the functions of working memory, and many answers remain to be revealed
.

For example: How would a relatively simple memory be transformed into a more complex set of information in working memory? However, with today's technology, people can access and capture working memory in unprecedented ways
.

This will help to analyze the commonalities and differences in working memory formation for various input of visual information from the world
.

Paper link: https:// Reference link: https://directorsblog.
nih.
gov /2022/04/12/human-brain-compresses-working-memories-into-low-res-summaries/