Table of Contents
Complete reconstruction of the Drosophila larval brain
The residual network is hidden inside
Conclusion
Home Technology peripherals AI There is really ResNet in the brain! The world's first 'Drosophila Brain Connectome' is released: it took more than ten years to reconstruct 3,000 neurons and more than 500,000 synapses!

There is really ResNet in the brain! The world's first 'Drosophila Brain Connectome' is released: it took more than ten years to reconstruct 3,000 neurons and more than 500,000 synapses!

Apr 11, 2023 pm 06:49 PM
Neural Networks Research

Although modern deep learning has long since broken away from imitating "biological neural networks", understanding the operating mechanism of the biological brain is still very helpful for the future development of neural network models.

The way brain circuits are structured affects the brain's computational capabilities, but so far, no specific structure of the brain has been seen except in some very simple organisms.

In November last year, researchers from many top institutions such as the University of Cambridge, Johns Hopkins University, and Janelia Research Park uploaded a paper on Biorxiv. AfterMore than ten years of arduous research, for the first timecompletelyreconstructed the brain connectome of the "Drosophila larvae".

There is really ResNet in the brain! The worlds first Drosophila Brain Connectome is released: it took more than ten years to reconstruct 3,000 neurons and more than 500,000 synapses!

Paper link: https://www.science.org/doi/10.1126/science. add9330​

On March 10, the relevant results were published in the magazine "Science".

Joshua Vogelstein, an associate professor at Johns Hopkins University and one of the authors of the paper, said that fruit flies are closer to the human brain in many ways than other organisms, and some areas correspond to decision-making. Some areas correspond to learning and some to navigation; and the brains of Drosophila larvae and humans are also divided into left and right sides.

In the analysis of the fruit fly brain, we can also find some results in modern neural networks, such as recurrent neural networks and shortcut paths between multi-layer networks (residual network ResNet) etc., may inspire improvements to machine learning models.

Good news: The reconstructed Drosophila larval brain connectome includes 3016 neurons

Bad news: People have 86 billion neurons.

Complete reconstruction of the Drosophila larval brain

The brain is mainly composed of "neuron cells". Adjacent neurons can interact with each other at the connections between synaptic cells. To send out a signal, one neuron releases a "neurotransmitter" and another neuron receives the chemical. The complete map of neurons and synapses in the brain is called the connectome. Studying the connectome is crucial to understanding how the brain generates behavior.

The main process for reconstructing the connectome is to cut the brain into ultra-thin (20 micron) slices, and then use the electron flow of an electron microscope to image the slices, such as

The brain of a fruit fly larvae the size of a grain of salt is cut into thousands of pieces. If anything goes wrong, you have to start all over again.

For some simple organisms, it is relatively easy to construct a complete connectome. The first fully drawn It is Caenorhabditis elegans, which has only 302 neurons in its entire body. It was mapped in the 1980s. There is really ResNet in the brain! The worlds first Drosophila Brain Connectome is released: it took more than ten years to reconstruct 3,000 neurons and more than 500,000 synapses!

But so far, researchers have mapped complete synaptic connectomes for only three organisms, each with just a few hundred brain neurons.

In 2020, researchers from Google and Janelia Research Park released a 3D model of the Drosophila brain connectome, containing 25,000 Drosophila neurons. Spanning different cell types and multiple brain regions, the model is not a complete brain, and even then it contains only a quarter of the 100,000 neurons in an adult fruit fly.

There are also some research groups that are studying the connectome of animals with larger brains, such as insects, fish, mammals, etc. However, due to the large number of neurons, the main research method is to partition the brain. Studying them in isolation results in the inability to reconstruct brain regions that are interconnected across space.

The complete connectome reconstructed this time belongs to the larvae of Drosophila melanogaster. Drosophila melanogaster can show a very rich behavior, including learning, value calculation and behavioral selection, and is closely related to adult fruit flies. Flies and larger insects have homologous brain structures.

There is really ResNet in the brain! The worlds first Drosophila Brain Connectome is released: it took more than ten years to reconstruct 3,000 neurons and more than 500,000 synapses!

Powerful genetic tools can be used to selectively manipulate or record individual neuron types in tractable In model systems, hypotheses about the functional roles of specific neuronal and circuit motifs revealed by the connectome can be readily tested.

The research team cut the brain of a "6-hour-old" Drosophila melanogaster larvae into 4841 pieces, scanned them with a high-resolution electron microscope, digitized the images and then reassembled them into a three-dimensional image; with the assistance of computer analysis, the final map generated contains 3,016 neurons and 548,000 synapses.

The residual network is hidden inside

The researchers conducted a detailed analysis of the brain circuit structure, including connections and neuron types, network hubs ) and neural circuit diagrams.

The majority (73%) of the brain's input and output hubs (in-out hubs) are "post-synaptic centers for learning centers" or "dopaminergic neurons that drive learning" "presynaptic center"; using graph spectral embedding technology, hierarchically clustered neurons based on synaptic connectivity were divided into 93 types that are internally consistent based on other characteristics such as morphology and function. sex.

The researchers also developed an algorithm to track signal propagation in the brain's multisynaptic pathways and analyzed feedforward (from sensory to output) and feedback pathways. , multisensory integration and cross-hemisphere interactions.

There is really ResNet in the brain! The worlds first Drosophila Brain Connectome is released: it took more than ten years to reconstruct 3,000 neurons and more than 500,000 synapses!

Extensive multisensory integration is found in the brain, and at varying depths from sensory neurons to output neurons Multiple interconnected pathways form a distributed processing network.

The brain has a highly recurrent structure, and 41% of neurons receive long-range cyclic input. However, the distribution of loops is not uniform, and the loop rate is particularly high in areas involving learning and action selection. high.

Dopaminergic neurons, which drive learning, are among the most common neurons in the brain.

There is really ResNet in the brain! The worlds first Drosophila Brain Connectome is released: it took more than ten years to reconstruct 3,000 neurons and more than 500,000 synapses!

There are many contralateral neurons that project to both hemispheres of the brain. They are in-and-out centers. out hubs), synapse with each other, promoting extensive communication between the two hemispheres of the brain; the article also analyzes the interaction between the brain and nerves.

Researchers found that descending neurons target a small group of premotor elements that play an important role in switching locomotor states. role.

Conclusion

The complete brain connectome of Drosophila larvae will provide the basis for other theoretical and experimental studies of brain function for a long time to come, generated in this study The methods and computational tools will facilitate future analysis of connectomes.

Although the details of brain organization vary across the animal kingdom, many neural circuit structures are conserved.

As the brain connectomes of more other organisms are mapped in the future, comparisons between different connectomes will reveal common and potentially optimal circuit structures, as well as between organisms. Behavioral differences in traits.

Some structural features observed in the Drosophila larval brain, including multi-layered shortcuts and prominent nested loops, can be The discovery of in advanced artificial neural networks may be able to make up for the current network's problems in depth and generalization of processing tasks. These features can also increase the brain's computing power and overcome the physiological limitations of the number of neurons.

Future analysis of the similarities and differences between the brain and artificial neural networks may help understand the brain's computational principles and may inspire new machine learning architectures.

Related comments

Gashpal Yekayi, professor of neuroscience at the University of Exeter, UK, said , all neurons of the brain are reconstructed and all neural connections analyzed. He called the work "very important."

Katherine Dulac of Harvard University in the United States believes that these data reveal the "deep logic" of these neuronal connections.

However, Scott Emmons of the Albert Einstein College of Medicine in New York believes that simply mapping synapses does not provide the full picture.

Emmons mapped the connectome of the male and female Caenorhabditis elegans in 2019. Neurons also communicate through slowly released chemicals such as hormones and other connections between cells. Connections (i.e. gap junctions) communicate with each other.

All this must be taken into account, but only synapses are included in the newly drawn connectome.

The above is the detailed content of There is really ResNet in the brain! The world's first 'Drosophila Brain Connectome' is released: it took more than ten years to reconstruct 3,000 neurons and more than 500,000 synapses!. For more information, please follow other related articles on the PHP Chinese website!

Statement of this Website
The content of this article is voluntarily contributed by netizens, and the copyright belongs to the original author. This site does not assume corresponding legal responsibility. If you find any content suspected of plagiarism or infringement, please contact admin@php.cn

Hot AI Tools

Undresser.AI Undress

Undresser.AI Undress

AI-powered app for creating realistic nude photos

AI Clothes Remover

AI Clothes Remover

Online AI tool for removing clothes from photos.

Undress AI Tool

Undress AI Tool

Undress images for free

Clothoff.io

Clothoff.io

AI clothes remover

AI Hentai Generator

AI Hentai Generator

Generate AI Hentai for free.

Hot Article

R.E.P.O. Energy Crystals Explained and What They Do (Yellow Crystal)
2 weeks ago By 尊渡假赌尊渡假赌尊渡假赌
Repo: How To Revive Teammates
1 months ago By 尊渡假赌尊渡假赌尊渡假赌
Hello Kitty Island Adventure: How To Get Giant Seeds
4 weeks ago By 尊渡假赌尊渡假赌尊渡假赌

Hot Tools

Notepad++7.3.1

Notepad++7.3.1

Easy-to-use and free code editor

SublimeText3 Chinese version

SublimeText3 Chinese version

Chinese version, very easy to use

Zend Studio 13.0.1

Zend Studio 13.0.1

Powerful PHP integrated development environment

Dreamweaver CS6

Dreamweaver CS6

Visual web development tools

SublimeText3 Mac version

SublimeText3 Mac version

God-level code editing software (SublimeText3)

YOLO is immortal! YOLOv9 is released: performance and speed SOTA~ YOLO is immortal! YOLOv9 is released: performance and speed SOTA~ Feb 26, 2024 am 11:31 AM

Today's deep learning methods focus on designing the most suitable objective function so that the model's prediction results are closest to the actual situation. At the same time, a suitable architecture must be designed to obtain sufficient information for prediction. Existing methods ignore the fact that when the input data undergoes layer-by-layer feature extraction and spatial transformation, a large amount of information will be lost. This article will delve into important issues when transmitting data through deep networks, namely information bottlenecks and reversible functions. Based on this, the concept of programmable gradient information (PGI) is proposed to cope with the various changes required by deep networks to achieve multi-objectives. PGI can provide complete input information for the target task to calculate the objective function, thereby obtaining reliable gradient information to update network weights. In addition, a new lightweight network framework is designed

A deep dive into models, data, and frameworks: an exhaustive 54-page review of efficient large language models A deep dive into models, data, and frameworks: an exhaustive 54-page review of efficient large language models Jan 14, 2024 pm 07:48 PM

Large-scale language models (LLMs) have demonstrated compelling capabilities in many important tasks, including natural language understanding, language generation, and complex reasoning, and have had a profound impact on society. However, these outstanding capabilities require significant training resources (shown in the left image) and long inference times (shown in the right image). Therefore, researchers need to develop effective technical means to solve their efficiency problems. In addition, as can be seen from the right side of the figure, some efficient LLMs (LanguageModels) such as Mistral-7B have been successfully used in the design and deployment of LLMs. These efficient LLMs can significantly reduce inference memory while maintaining similar accuracy to LLaMA1-33B

'The owner of Bilibili UP successfully created the world's first redstone-based neural network, which caused a sensation on social media and was praised by Yann LeCun.' 'The owner of Bilibili UP successfully created the world's first redstone-based neural network, which caused a sensation on social media and was praised by Yann LeCun.' May 07, 2023 pm 10:58 PM

In Minecraft, redstone is a very important item. It is a unique material in the game. Switches, redstone torches, and redstone blocks can provide electricity-like energy to wires or objects. Redstone circuits can be used to build structures for you to control or activate other machinery. They themselves can be designed to respond to manual activation by players, or they can repeatedly output signals or respond to changes caused by non-players, such as creature movement and items. Falling, plant growth, day and night, and more. Therefore, in my world, redstone can control extremely many types of machinery, ranging from simple machinery such as automatic doors, light switches and strobe power supplies, to huge elevators, automatic farms, small game platforms and even in-game machines. built computer. Recently, B station UP main @

Multi-path, multi-domain, all-inclusive! Google AI releases multi-domain learning general model MDL Multi-path, multi-domain, all-inclusive! Google AI releases multi-domain learning general model MDL May 28, 2023 pm 02:12 PM

Deep learning models for vision tasks (such as image classification) are usually trained end-to-end with data from a single visual domain (such as natural images or computer-generated images). Generally, an application that completes vision tasks for multiple domains needs to build multiple models for each separate domain and train them independently. Data is not shared between different domains. During inference, each model will handle a specific domain. input data. Even if they are oriented to different fields, some features of the early layers between these models are similar, so joint training of these models is more efficient. This reduces latency and power consumption, and reduces the memory cost of storing each model parameter. This approach is called multi-domain learning (MDL). In addition, MDL models can also outperform single

Crushing H100, Nvidia's next-generation GPU is revealed! The first 3nm multi-chip module design, unveiled in 2024 Crushing H100, Nvidia's next-generation GPU is revealed! The first 3nm multi-chip module design, unveiled in 2024 Sep 30, 2023 pm 12:49 PM

3nm process, performance surpasses H100! Recently, foreign media DigiTimes broke the news that Nvidia is developing the next-generation GPU, the B100, code-named "Blackwell". It is said that as a product for artificial intelligence (AI) and high-performance computing (HPC) applications, the B100 will use TSMC's 3nm process process, as well as more complex multi-chip module (MCM) design, and will appear in the fourth quarter of 2024. For Nvidia, which monopolizes more than 80% of the artificial intelligence GPU market, it can use the B100 to strike while the iron is hot and further attack challengers such as AMD and Intel in this wave of AI deployment. According to NVIDIA estimates, by 2027, the output value of this field is expected to reach approximately

1.3ms takes 1.3ms! Tsinghua's latest open source mobile neural network architecture RepViT 1.3ms takes 1.3ms! Tsinghua's latest open source mobile neural network architecture RepViT Mar 11, 2024 pm 12:07 PM

Paper address: https://arxiv.org/abs/2307.09283 Code address: https://github.com/THU-MIG/RepViTRepViT performs well in the mobile ViT architecture and shows significant advantages. Next, we explore the contributions of this study. It is mentioned in the article that lightweight ViTs generally perform better than lightweight CNNs on visual tasks, mainly due to their multi-head self-attention module (MSHA) that allows the model to learn global representations. However, the architectural differences between lightweight ViTs and lightweight CNNs have not been fully studied. In this study, the authors integrated lightweight ViTs into the effective

The most comprehensive review of multimodal large models is here! 7 Microsoft researchers cooperated vigorously, 5 major themes, 119 pages of document The most comprehensive review of multimodal large models is here! 7 Microsoft researchers cooperated vigorously, 5 major themes, 119 pages of document Sep 25, 2023 pm 04:49 PM

The most comprehensive review of multimodal large models is here! Written by 7 Chinese researchers at Microsoft, it has 119 pages. It starts from two types of multi-modal large model research directions that have been completed and are still at the forefront, and comprehensively summarizes five specific research topics: visual understanding and visual generation. The multi-modal large-model multi-modal agent supported by the unified visual model LLM focuses on a phenomenon: the multi-modal basic model has moved from specialized to universal. Ps. This is why the author directly drew an image of Doraemon at the beginning of the paper. Who should read this review (report)? In the original words of Microsoft: As long as you are interested in learning the basic knowledge and latest progress of multi-modal basic models, whether you are a professional researcher or a student, this content is very suitable for you to come together.

Do you know that programmers will be in decline in a few years? Do you know that programmers will be in decline in a few years? Nov 08, 2023 am 11:17 AM

"ComputerWorld" magazine once wrote an article saying that "programming will disappear by 1960" because IBM developed a new language FORTRAN, which allows engineers to write the mathematical formulas they need and then submit them. Give the computer a run, so programming ends. A few years later, we heard a new saying: any business person can use business terms to describe their problems and tell the computer what to do. Using this programming language called COBOL, companies no longer need programmers. . Later, it is said that IBM developed a new programming language called RPG that allows employees to fill in forms and generate reports, so most of the company's programming needs can be completed through it.

See all articles