On July 16, Google AI published a blog stating that the Google Research department and Max Planck Institute have cooperated to propose a new type of recurrent neural network, which can improve the accuracy of automatic analysis of connected group data, compared with the previous depth The accuracy of learning technology is an order of magnitude improvement.
Jeff Dean also reposted his comment on Twitter, saying, "Research on the connectivity of real neurons using artificial neural networks is really cool."
â–ŒOrders of magnitude improved the accuracy of automatic analysis of connection group data
According to the definition of Wikipedia, Connectomics (Connectomics) to draw and study neural connections (connectomes) is a complete circuit diagram depicting the connection mode of the organism's nervous system (especially the brain and eyes). Because these structures are extremely complex, efficient screening neuroimaging and histological methods are used to improve the speed, efficiency, and accuracy of drawing neural connections.
The purpose of connectionomics is to comprehensively map the structure of neuronal networks in the nervous system in order to better understand how the brain functions. This process requires 3D imaging of brain tissue at nanometer resolution (usually using an electron microscope), and then analyzing the resulting image data to track the brain's axons and identify individual synaptic connections.
Due to the high resolution of imaging, even 1 cubic millimeter of brain tissue can generate more than 1000 TB of data. Coupled with the subtle and complex structures in these images, the main bottleneck of brain mapping is automatically analyzing these data, rather than acquiring the data itself.
The article "High-Precision Automated Reconstruction of Neurons with Flood-Filling Networks" published by Google and Max Plank Institute of Neurobiology in "Nature Methods" shows a How the new recurrent neural network can improve the accuracy of automatic analysis of connected group data is an order of magnitude improvement over previous deep learning techniques.
â–ŒUse flood filling network for 3D image segmentation
Tracking axons in large-scale electron microscope data is an image segmentation problem. Traditional algorithms divide the process into at least two steps: use edge detectors or machine learning classifiers to find the boundaries between axons, and then use algorithms such as watershd or graph cutting to remove image pixels that are not separated by boundaries. Make combinations.
In 2015, they began to try an alternative method based on recurrent neural networks to unify these two steps. The algorithm sows seeds at a specific pixel location, and then uses a cyclic convolutional neural network to continuously "fill in" an area. The network predicts which pixels belong to the same part of the specific pixel. Since 2015, they have been committed to applying this new method to large-scale connection group datasets and rigorously quantifying its accuracy.
Flood fill network of segmented objects. The yellow dot is the center of the current focus area; when the algorithm continuously checks the entire image, the segmented area will continue to expand (blue).
â–ŒPass the expected running length measurement accuracy
They devised a metric called "Expected Run Length" (ERL): Given random points within random neurons in a three-dimensional image of the brain, they can track the distance of neurons before making mistakes. This is an example of Mean Time Between Failures (Mean Time Between Failures), but in this case the space between failures is detected instead of time.
For researchers, the appeal of ERL is that it connects the linear physical path length with the frequency of each error generated by the algorithm and can be directly calculated. For biologists, the specific value of ERL can be related to biologically relevant numbers, such as the average path length of neurons in different parts of the nervous system.
The blue line represents the expected run length (ERL) result. The red line represents the "merging rate", which refers to the frequency at which two separate axons are mistakenly tracked as a single object; a very low merge rate is important for manual identification and correction of remaining errors in reconstruction.
â–ŒConnectomics of songbirds
They used ERL to measure the real neuron data set of the zebra finch brain at 1 million cubic microns and found that this method performed better than other deep learning methods applied to the same data set.
Algorithm is tracking individual axons in zebra finch brain
They used a new flood-filling network method to segment a small part of the neurons in the zebra finch brain and reconstruct a part of the zebra finch brain. Different colors represent different objects of the segmentation automatically generated using the filling irrigation network. The golden ball represents the synapse position automatically identified using the previous method.
They will continue to improve the connectionomics reconstruction technology. In order to help more research groups develop connectionomics technology, they developed TensorFlow code for flooding and filling the network method, and developed Web GL visualization software for 3D data sets to help understand and improve the reconstruction results.
Some client may feel that 10th Laptop is little old, so prefer 11th Laptop or 12th laptop. However, you will the 10th cpu is even more powerful than 11th, but price is nearly no difference, especially you take in lot, like 1000pcs. As a professional manufacturing store, you can see Laptop i3 10th generation 8gb ram,i5 laptop 10th generation, intel i7 10th gen laptop, etc. It`s a really tough job selecting a right one on the too many choices. Here are some tips, hope help you do that easier. Firstly, ask yourself what jobs you mainly need this Gaming Laptop to do. Secondly, what special features you care more? Like fingerprint, backlight keyboard, webcam rj45, bigger battery, large screen, video graphics, etc. Finally, ask the budget you plan to buy gaming laptop. Thus you will get the idea which laptop is right one for you.
Except integrated laptop, also have graphic laptop with 2gb or 4gb video graphic, so you can feel freely to contact us anytime and share your basic requirements, like size, cpu, ram, rom, video graphics, quantity,etc. More detailed value information provided in 1-2 working days for you.
10th Laptop,Laptop I3 10th Generation 8gb Ram,I5 Laptop 10th Generation,10th Generation Laptop,Intel I7 10th Gen Laptop
Henan Shuyi Electronics Co., Ltd. , https://www.shuyitablet.com