多彩な 英語 講師陣から検索…
投稿
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.
2016年1月27日
5
0
Dear friends! I need your help. Can you check my text, please)))) Our age is called, “digital age”. Technologies are being developed almost every day. There is no sphere, where people have not tried to understand the laws of nature or improve something there. Of course, one of the most interesting thing to understand and make a copy is the human mind. The artificial neural network is a mathematical model of human mind. Now this sphere is growing fast, many models and algorithms have been created. The convolutional neural network is a kind of it. What is the convolutional neural network? To understand this, you should know what the artificial neural network (ANN) is. An artificial neural network is a (large) collection of very simple, massively interconnected cells, also known as neurons. Most ANN-architectures are grouped by the term multi-layer feed-forward ANNs. It means that data goes in one direction – from input to output, and neurons are grouped into layers. Then we will consider their structure. The first layer is usually the input layer, the last layer is the output one. All layers are called hidden layers. Every cell performs very simple calculation: it sums all its inputs multiplied by their respective weights (the weight is a characteristic of the connection between two cells), after which a squashing function is applied to the result. The squashing function is usually called transfer or activation function. After that, cells transfer data to the next layer. The output of the network is the output of the last layer (output layer). The weights of the connection can be preset by an operator, or trained by presenting the ANN with the examples of the input (possibly together with the desired output). There are many training methods; one of the most popular is the back-propagation method. This method is based on next idea. Neurons can send error signals in reverse direction – from output to input. In the case of error we firstly correct weights on the last layer, then, based on this results, calculate new weights for the previous layer, and we repeat while the current layer is not input. Now we can move to convolutional networks. But before that we must define what is convolution. Convolution is a mathematical operator (denoted by “*”) on two functions, defined as the sum of the products of the function values, with the one function mirrored in an origin. But we will use next interpretation of this definition.
2016年1月26日
5
0
Please help me! Check this text please, especially grammar! Structure of the Cloud The primary purpose of cloud storage is to deliver instant memory storage to a user on demand in a flexible way. The architecture of Cloud memory is layered: the first layer is known as the front end, it supplies users with a way to access the cloud, by exporting the API, usually the SCSI protocol. Then, the middle layer of the Cloud is an algorithm which handles where to place the information. The final layer is the physical servers and computers that hold the information. Cloud computing is an abstraction of the traditional client-server model. Instead of owning one's own data servers, a user leases whatever amount of space he/she needs from a provider. This is an improvement because with the cloud you get exactly the amount of space you need, and only have to pay for that. This is a big advantage for businesses because they can pay for five machines during working hours, and drop down to one during off hours, as opposed to managing their own five servers 100% of the time. The key difference is that someone else is managing the storage, and willing to lease whatever you need on demand. In comparison the traditional client-server model is like buying a property, while cloud is like renting, sometimes when you are going to be using the resource a lot, and it is important to protect it is a good idea to buy (like home), in other cases when you're going to be sharing the resource a lot, and not always using it, renting is the better option (like office space). Cloud Performance The main factor in cloud performance is the transfer of information from the service provider to the user. The work for processing and storing data can be delegated to external hardware usually belonging to the service provider. Therefore, there is no need for the user to have the processing power locally to have good performance. The only thing a user needs to ensure fast service is reliable and fast network access.
2015年12月13日
0
0
Please check this text! Especially GRAMMAR! Structure of the Cloud The primary purpose of cloud storage is to deliver instant memory storage to a user on demand in a flexible way. The architecture of Cloud memory is layered: the first layer is known as the front end, it supplies users with a way to access the cloud, by exporting the API, usually the SCSI protocol. Then, the middle layer of the Cloud is an algorithm which handles where to place the information. The final layer is the physical servers and computers that hold the information. Cloud computing is an abstraction of the traditional client-server model. Instead of owning one's own data servers, a user leases whatever amount of space he/she needs from a provider. This is an improvement because with the cloud you get exactly the amount of space you need, and only have to pay for that. This is a big advantage for businesses because they can pay for five machines during working hours, and drop down to one during off hours, as opposed to managing their own five servers 100% of the time. The key difference is that someone else is managing the storage, and willing to lease whatever you need on demand. In comparison the traditional client-server model is like buying a property, while cloud is like renting, sometimes when you are going to be using the resource a lot, and it is important to protect it is a good idea to buy (like home), in other cases when you're going to be sharing the resource a lot, and not always using it, renting is the better option (like office space). Cloud Performance The main factor in cloud performance is the transfer of information from the service provider to the user. The work for processing and storing data can be delegated to external hardware usually belonging to the service provider. Therefore, there is no need for the user to have the processing power locally to have good performance. The only thing a user needs to ensure fast service is reliable and fast network access.
2015年12月6日
0
0
Please check this text! Especially GRAMMAR! For some computer owners, finding enough storage space to hold all the data they have acquired is a real challenge. Some people invest in larger hard drives. Others prefer external storage devices like thumb drives or compact discs. Desperate computer owners might delete entire folders worth of old files to make space for new information. But some are choosing to rely on a growing trend: cloud storage. While cloud storage sounds like it has something to do with weather fronts and storm systems, it really refers to saving data to an off-site storage system maintained by a third party. Instead of storing information to your computer's hard drive or another local storage device, you save it to a remote database. The Internet provides the connection between your computer and the database. On the surface, cloud storage has several advantages over traditional data storage. For example, if you store your data on a cloud storage system, you will be able to get to that data from any location that has Internet access. You would not need to carry around a physical storage device or use the same computer to save and retrieve your information. With the right storage system, you could even allow other people to access the data, turning a personal project into a collaborative effort. Pros and Cons of Cloud Storage For a personal user, the most important advantage of cloud storage is convenience. With the increased usage of portable devices such as smartphones and tablets, access to cloud storage can be beneficial. It eliminates the need to manually transfer files from one device to another. Another important advantage is the added protection against file corruption because data are stored in multiple servers around the globe. One disadvantage though is that your files might be vulnerable to hacking. This one concern is what makes a number of people a bit mistrusting of using cloud storage for their more sensitive data. Another disadvantage is that should the cloud storage provider cease to exist, all your files will be lost. But the same goes for everything in life, there will always be some sort of risk. Educating yourself and weighing the pros and cons should help you decide whether the use of cloud storage is for you. Structure of the Cloud
2015年12月6日
1
0
Дорогие друзья помогите пожалуйста перевести. Dear friends, please help to translate on Russian. I speak English not very bad, but these sentences are difficult even for me. 1) A police spokeswoman said: "We are trying to facilitate lawful protest but if people are going to break the law, we are going to take action." 2) Five crews battled the flames for more than six hours before bringing them under control. 3) Police confirmed last night that they were still looking for four people, but could not rule out the possibility that visitors to the hotel might have been trapped inside.
2014年11月30日
9
0
Dear friends, please check this task. It's from HSK exam. A 我从您 这儿学了很多. B 新年快乐! C 我弟弟在北京大学读书, 今年是第三年了. D 听说他病了, 我们找时间去看看他把. E 这个是下午送? 送到哪儿? 1. 今天是1月1日. 2. 等以下,我打电话问以下. 3. 他会说汉语? 4. 星期六怎么样? 5. 希望我的回答能对你有帮助. There are my variants: 1 B 2 E 3 C 4 D 5 A
2014年4月2日
8
0