Neural Networks are becoming easy to use

If you want more articles on artificial intelligence is, in most cases, "the neural network." Are the words used to model an artificial neural network in the human brain, if this is the first time, so that a computer can learn from the data entry you go.

In addition, the effectiveness of this powerful branch of the machine learning too than anything, to lead a new era of artificial intelligence, has become a cause to end the "winter of AI" long life. Simply put, the neural network is more fundamentally today's disruptive technology.

This neural network guide is intended to capture the level of deep learning conversation. For this, we count as much as possible analogy and animation, not plunge into mathematics.

An AI start school in order to understand the data, if you give as many instructions as possible, is as much information as possible to powerful computers that are not directly taught to be able to follow "Think" loading. Such a famous chess computer Deep Blue from IBM: all possible chess program full moves to the computer, to provide enough power to execute the well-known strategy, IBM's future programmer theory anticipation, all possible moves, above results, Overrides party to select the order of the subsequent move. This is because as I learned chess master in 1997, so it is really operational.

In this type of computation, engineers are based on certain preprogrammed rules. If such a thing happened, please do so. So, I'm not into all human-style, I would like to know a flexible learning. It's a powerful supercomputer, but it's certainly not the same in "thinking".

Teaching machines to learn

For the past ten years, scientists reviviram old concepts that rely on banks of huge encyclopedia memories, but do not analyze data entered into models into human thought in a loosely simple and systematic way. Known as deep learning and neural network, this technology has been around since the 1940's, the reason is that it is growing exponentially with today's data - photos, videos, sounds Finally, along with searching, browsing habits and more - processor supercharging and affordable, you can get to know their real possibilities.
Machine - they are just like us!

Artificial Neural Network (human opposition) is the construction of an algorithm that allows you to learn everything from image recognition and curation playlist machine commands and audio commands. A typical ANN consists of forming connections for tens of millions of interconnected neurons, stacked and sequenced on a polygonal line known as a layer. In many cases, these layers have only one layer of pre-neurons and are interconnected after these inputs and outputs. (This is quite different from human brain neurons interconnected in all directions.)

This multi level of ANN learns the machine today, cultivates it with a lot of marked data and interprets these data (and sometimes is good) You can learn how to Lord other than human It is one way.

This non-literal method is known as neural convolution network (CNN) image recognition, based on type, as he is using a mathematical process known as convolution for analyzing images It is known as the so-called "specific" of a neural network, making identification of objects is partially obscured or only displayed from a specific angle. (There are other types of neural networks including recurrent neural networks and neural network feedforwards, but these are less useful for identifying things like images, examples of use).
On all networks

So, how do you learn the neural network? Let's take a super-simple but called super effective, called supervised learning called. Here, the neural network sends a lot of signaling training data, by human, so that the neural network can be seen as intrinsic learning.

Each of these labeled data is said to consist of an image of an apple and an orange, respectively. image data. "Apple" and "Orange" are labels, depending on the image. As the images are inserted, the network splits them into their most basic components, ie borders, textures and shapes. As the images spread through the network, these basic components, when combined, can be used to create different or different curves and color rods, whole orange, or both green and red apples, similar or more abstract It is combined to form a concept.

At the end of this process, the network will try to predict what is in the image. Initially, these predictions are displayed as random guesses if real learning has not yet occurred. If Apple has "orange", in the case of input images, you need to adjust the layers inside the network.

Adjustments are made through a process called backpropagation to increase the probability of predicting "apples" for the same image next time. Forecasts are more or less accurate and will occur over and over until it is not being improved. As educational children identify apples and oranges in real life, parents as well will be perfect for computers. In your mind, you thought that "you can hear like learning," you can have an AI career.