There are a lot of models that scientists tried through which simulate the way of human thinking, in order to create intelligent industrial approaching than that of biological, we can mention examples of these models: Bayes networks, trees and resolution, algorithms and deep education, and finally neural industrial networks, and each of these models, However, artificial neuron networks are the most popular, the closest and most prevalent models of the human brain.
Artificial neural networks have begun to be thought of as a similar model in the way they work for the brain. Other researchers do so as long as artificial neural networks can solve many of the problems they face.
We recalled that the industrial network is inspired by the real one; therefore, an introduction to the functioning of the brain and its neurons is essential if we need a deeper understanding of artificial neural networks.
The human brain is a block of mystery!
We humans discovered most of the blue planet we inhabited, analyzed most of its mysteries, invaded space and visited the moon, and our visit to Mars is just around the corner. The little one is considered by many to be the most complex thing in the whole universe. Where does this complexity come from?
The complexity comes from several factors. The structure of the brain consists of a homogeneous mass of most cells called neurons (neurons). These neurons are interconnected at points called “synapses”, the purpose of which is to send and receive electrical or chemical signals.
The number of these neurons in the human brain exceeds one hundred billion neurons, but the interconnections between them are much more than this number and given a large amount of complexity, scientists stand bewildered before him as we said, and are not fully aware of the way the brain works, is what makes man capable of Conclusion and perception is the number of large neurons or the number of associations between them? Or is the speed of treatment the cause or method of treatment?
The number of neurons is probably not the decisive factor, especially if we know that the number of neurons in the brain of an elephant and whale exceeds the number in the human brain and the number reaches 200 billion neurons and that computers are thousands of times faster than neurons, and therefore some scientists believe that the method of neurons Natural data processing is the key to uncovering this ambiguity. But what concerns us today is not exactly understanding the work of the brain, as much as it is about creating a simplified model for solving some complex problems.
Method of building the neural network
Artificial intelligence scientists attempted to model neurons in the brain, simulating their behavior as individual elements through computer programs, and then developing techniques to connect these individual neurons and study the results. These connections between neurons are usually achieved by specific weights.
The method of organizing neurons in industrial networks is usually in a series of layers, starting with a lower layer and ending with a top layer and several hidden layers, and these neurons in each layer are associated with the neurons of the lower and higher layer. The lower layer receives inputs from outside the neural network, for example, pixels from an image, and the neurons in the hidden layers receive their input from the lower layers and reach the higher layers. In the end, the top layer gives us the output of the network as a whole.
After construction comes training
After building a neural industrial network we turn to the training stage, and this stage is considered a special case of machine learning, identify the concept of machine learning.
Marking the neural network, in short, means synthesizing the weights of the connections between the neurons until we get a certain desired output. For example, we want to train a neural network to detect images that contain a cat. The network adjusts the weights and connections to give us a picture of a cat in the output.
The neuronal network can be trained in two methods, the first is “supervised education” and the second “supervised education”. The “supervised education” method relies on the presentation of images containing cats and the other does not contain, with reference to images containing cats, while “education without supervision” depends on Show pictures of cats and leave the rest of the network to identify themselves on existing cats by searching for special features that depend on them.
An important question may arise for anyone, is it reasonable for the neural network to recognize cats themselves? Yes, it is possible, by looking for specific traits or patterns that distinguish cats from others, such as the distinctive face shape of cats, the shape of her hair, her hands, ears, and so on. These images are tilted or half blocked. After training in pictures of millions, perhaps, the network can detect images of cats you have never seen before.
To approximate the way we adjust the neural network to its weights, imagine you have a stick and want to adjust the strings of that lute to get the desired sound, so what to do? You will immediately adjust the strings of this lute by experimenting, tightening the string or reflecting the process and hearing the resulting sound, and then adjusting according to what you have heard and so on until you have finished all the strings, thus adjusting the lute according to the desired result. The neural network does not do anything else. It adjusts the weights, measures the output and then adjusts the weights until you get the desired final output.
The process of controlling the oud depends heavily on the initial state of the tendons, as well as the amount by which we change the state of the tendon each time, as well as applies to the neural network. These are always trying to find the best primitive conditions (weights of primitive joints), as well as making these weights reach the desired output as soon as possible.
When we build a strong neural network and teach it well, we have given artificial intelligence a brain. Although this brain is by no means equivalent to the human brain, artificial intelligence has often overtaken humans, which is due to its ability to process A lot of data in a short time, and for his part, man still excels in many other things such as understanding feelings, and between us and artificial intelligence is still a long conflict, for whom will prevail?