Artificial neural networks
Artificial neural networks
Lecturer-ABDUL STTAR ISMAIL WDAA-Department of Mathematics
Neural networks are one of the most important areas of control engineering and artificial intelligence, which reflects a significant and tangible development in the way of human thinking. The idea of neural networks revolves around simulating the human mind using a computer. The foreseeable development in this field is due to many studies carried out in the field of Processing Neural, and the simulation process is done by solving the problems encountered, by following the self-learning processes that depend on the experiences stored in the network, which achieve the best results. The method of artificial neural networks has been borrowed from biological neural networks, thanks to its entry into the global business circle by Pitts & Cultch - Mc about 60 years ago. Definition of ANN Neural Networks and their Components: They are computational techniques designed to simulate the way the human brain performs a particular task, through massive processing distributed in parallel, and made up of simple processing units. It stores practical knowledge and empirical information to make it available to the user by adjusting the weights.
So the ANN is similar to the human brain in that it acquires knowledge by training and stores this knowledge using connections within neurons called synaptic weights. There is also a neurobiological similarity, giving biologists the opportunity to rely on ANN to understand the evolution of biological phenomena. Just as a person has input units that connect him to the outside world, which are his five senses, so also neural networks need input units, and processing units in which arithmetic operations are carried out by which weights are set and through which we obtain the appropriate reaction for each of the inputs to the network. The input units make up a layer called the input layer, and the processing units make up the processing layer, which is the output of the network. And between each of these layers there is a layer of interfaces that connect each layer to the next layer, in which the weights of each interface are adjusted, and the network contains only one layer of input units, but it may contain more than one layer of processing layers. Transformation functions represent the logic circuits and the exponential functions in the neuron, which limit the output of the neuron according to the logical conditions in the known logic circuits, and they must have the following properties: • To be a continuous follower. • It should be derivable and its derivative is easy to calculate. • It should be streamlined, not decreasing. There are three types of activation methods: 1. Threshold function or step function: This function limits the output of the neuron so that the output becomes equal to one if the input is greater or equal to zero and the output becomes equal to zero if the input is less than the row. 2. Linear linear function or congruence function: it is always expressed as a linear function and this function is used in neurons used in linear adaptive filters. 3. The Exponential Function: It uses the logarithmic function in he calculation, and the input is confined between ∞ - and ∞ +, and the output is confined between 0 and 1, which is the most frequently used function. Types of Artificial Neural The artificial neural network is organized into several different bodies, where neurons are connected in several different ways, including: 1. Networks Neural Forward Feed: These are networks whose structure is devoid of a closed loop of interconnections between its component units. These networks are among the most widely used neural networks, as the network consists of at least two layers, and there are often hidden layers between the input layer and the layer. Output, and arithmetic operations move in one direction forward from the input layer to the output layer through the hidden layers 2. Networks Neural Back Feed: They are networks whose outputs find their way back again to become inputs in order to give the best possible results. 3. Networks Neural Associative Auto: They are networks in which all their constituent elements play an exemplary role, represented in receiving inputs and broadcasting outputs at the same time. The characteristics of artificial neural networks Neural networks have many characteristics, the most important of which are: 1. It is based on a strong mathematical foundation. 2. It represents one of the applications of information automation technology that is based on simulating the human mind. 3. Accept any kind of quantitative or qualitative data. 4. It has the ability to store the knowledge gained through the cases that are running on the network. 5. It can be applied in many different scientific fields Applications and uses of artificial neural networks The applications of artificial neural networks are many and important in the field of computers, especially in building games, so that game programmers can make an little.