ابحث بين معلمي الإنجليزية المتعددين...
Russia flagAnna
Dear friends, please check this text. This is the second part about convolutional neural networks. Image a square window with mesh, which contains the function values – filter. (then the pic with the example in my document). This window slides on processing image. It calculates the sum of pixel values with respective coefficient from this. We can write the result to the middle pixel. One of the most famous scientist who works with convolutional neural networks is Yann Le Cun (idea belongs to him). Let’s view his ANN’s. The input of convolution ANN (CNN) is an image, which is often preprocessed. The second layer contains the so-called feature maps. Each cell in such a feature map contains a set of incoming weights; this set is convolution filter or template. The place of the match in the input image corresponds to the place of cell in the feature map. This is what is meant by receptive fields: cells react to the same pattern at different positions in the input image. The next layer is a sub-sampling layer. This layer is included mainly to reduce the number of free parameters (which includes different connections) and has the same principle as for feature maps. The forth layer is hidden and fully connected with the previous one. Output layer is completely connected with hidden layers and forms the result of working. It consists 10 cells, the cell with the highest activation is chosen as a result. Researchers can use ANN that differs from Le Cun’s in details, but the principle is the same.CNNs have some advantages than classic ANNs. Their architecture has fewer weights than ANN one, because CNN uses one filter with weights for the whole image while ANN has a weight for each pixel. That makes CNN generalize results instead of storing information per pixel. Next advantage is good parallelization (executing the command in more than one threads, may be on different processor units). The disadvantage is application only for image recognition due to its architecture. But in some subjects CNNs, as we’ll see, are quite applicable. Why do we need convolutional neural networks? They are mainly used in image recognition. They have one of the best recognition and image classifying algorithm with an error rate of 0.23 percent. Next application is video analysis. The video is more complex than images since it has another (temporal) dimension. The common way is to fuse the features of different convolutional neural networks, which are responsible for the spatial and temporal stream.

التصحيحات · 5

0

Why do we need convolutional neural networks? They are mainly used in image recognition, and they have one of the best recognition and image classifying algorithms with an error rate of only 0.23 percent. 

The next application is in video analysis. The Video analysis is more complex than image analysis since it has another (temporal) dimension (time). The common way to address this is to fuse the features of different convolutional neural networks., which These additional features are responsible for the spatial and temporal streams.


***Overall, not bad. Let me know if you have any questions regarding any of the corrections, and I'll try my best to explain them.

0

The next layer is a the sub-sampling layer. This layer is included mainly to reduce the number of free parameters (which includes different connections), and has the same principles as for feature maps. The fourth layer is hidden and fully connected with the previous one. The output layer is completely connected with to the hidden layers, and forms the results of working the work. It consists of 10 cells, and the cell with the highest activation value is chosen as a result. Researchers can use ANNs that differs from Le Cun’s in details slightly, but the principle is will remain the same. CNNs have some advantages than over classic ANNs. Their architecture has fewer weights than an ANN one, model. This is because CNN uses one filter with weights for the whole image, while whereas ANN has a uses weights for each pixel. This means that makes CNNs will generalise the results instead of storing information per for each pixel. 

The next advantage that CNNs have is good better parallelization (i.e. executing the same command in more than one multiple threads, may be on potentially in different processor units).
The disadvantage of CNNs is in their limited applications only for in image recognition due to its their architecture. But, in some subjects CNNs, as we’ll see, CNNs are quite applicable.

*** Continued in third and final correction ***

0

Dear friends, please check this text. This is the second part about convolutional neural networks.

*** From the previous section, now that I have context I can adjust it.*** But Given this, we will use the next following interpretation of this definition.

Imagine a square window with mesh, which contains the function value s – filter. (then the pic with the example in my document).
Onto this window, slides on processing the image to be processed. [*I hope I understood this correctly. This sounds awkward to me, but I tried to leave your phrasing intact.] It This window then calculates the sum of the pixel values with as well as the respective coefficient. from this. Now, we can write store the result to in the middle pixel. One of the most famous scientists, who works with convolutional neural networks, is Yann Le Cun (this idea belongs to was developed by him). Let’s view his ANN’s. The input of a convolutional ANN (CNN) is an image, which is often typically preprocessed. The second layer contains the so-called feature maps. Each cell in such a one of these feature maps contains a set of incoming weights; this set is known as the convolution filter, or template. The place location of the match [Map? There is no reference to a "match" earlier in the passage, so I'm not certain what this is referring to] in the input image corresponds to the place location of the cell in the feature map. This is what is meant by receptive fields: cells react to the same pattern at different positions locations in the input image.
*** Once again, this is a bit too long for me to correct in one go. I'll finish the rest in another correction later. I have a prior engagement at the moment, so I'll come back and correct the rest in a few hours.

For reference, things highlighted in yellow are additions/modifications to your passage that weren't there before, but are necessary for the text to be correct and easily understood. Things highlighted in orange are stylistic additions of mine, typically related to word choice, that aren't necessary but sound better in an essay or paper. 

Hope this helps! I'll pick up where I left off later. :-)

0
The end of the text. Now image recognition is an actual theme. Convolutional neural networks are very useful for people because in our age we often deal with different images. That’s why this sphere is growing fast.

هل تريد التطور بشكل أسرع؟

انضم لمجتمع التعلّم هذا وجرّب التمرينات المجانية!

لا تفوّت فرصة تعلّم لغة جديدة وأنت مرتاح في منزلك. تصفّح مجموعتنا المختارة من مدرّسي اللغات ذوي الخبرة وسجّل في درسك الأول الآن!

ابدأ التعلم اليوم