Artificial Neural Network
Artificial Neural Networks
The artificial neural network is a practical way to learn various functions such as functions with real values, functions with discrete values, and functions with vector values. Artificial neural networks are one of the most important prerequisites for artificial intelligence. The study of artificial neural networks is largely inspired by natural learning systems. Moreover, They involve a complex set of interconnected neurons involved in learning.
Additionally, the human brain is thought to be made up of 1011 neurons, each of which is associated with approximately 104 other neurons. The switching speed of neurons is about 10-3 seconds Which is very little compared to 10-11 seconds compared to computers. However, a person is able to recognize a human image in 0.1 seconds. However, it takes minutes for the computer to do this recognition.
Perhaps it is not bad to think about this question first, why even though computer neurons switch faster than the human brain, humans remember a person’s face faster?
Neural network structure
See both images above. What similarities do you see?
As you can see, the first image is of a natural biological neuron. Moreover, information enters neurons through input or dendrites. Moreover, the same inputs are visible in the second image with values (x1, ……., Xm). Additionally, in the artificial neural network model, we assign a weight (w1, ……., Wm) to each input. Moreover, these weights are actually the importance of inputs to us. That is, the heavier the weight, the more important the input for network training. Then, you can add all the inputs together (Σ) and enter the axon as a single layer. Then, in the next step, we apply the Activation Function to the data. Moreover, the activation Function is actually defined according to the need of the problem and the type of our neural network. In addition, this function contains a mathematical formula for updating weights in the network.
After performing the calculations in this step, our information enters the other neuron through the output synapses, and this step continues until the so-called network is trained.
Applications of ANN
Computer science
- Classification of documents and information in computer networks and the Internet
- Development of monitoring software and antivirus software
Technical and engineering sciences:
- Reverse engineering and systems modeling,
- Electric charge consumption forecast
- Troubleshooting industrial and technical systems
- Design of various control systems
- Design, and optimization of technical and engineering systems
- Optimal decision making in engineering projects
Basic Science and Astronomy
- Predicting test results,
- Assess and estimate the correctness of hypotheses and theories
Medical Sciences
- Modeling of biomedical processes,
- Predicting the results of treatment and surgery
Experimental and biological sciences
- Classification of findings from experimental observations
Economics and Finance
- Classification of stock signs and symbols
Social Sciences and Psychology
- Classification and clustering of individuals and groups
Art and literature
- Classify and explore literary texts
Military Sciences
- predict the behavior of the attacking force and the enemy
- Targeting and chasing in missile weapons
Types of artificial neural networks
- Multilayer perceptron or MLP
- Radial neural networks or RBF
- Support vector machines or SVM
- Self-organizing maps or SOM
- Vector Digit Maker or LVQ
- Hopfield neural network
