ARTIFICIAL NEURAL NETWORKS
Need help with a related project topic or New topic? Send Us Your Topic
DOWNLOAD THE COMPLETE PROJECT MATERIAL
ARTIFICIAL NEURAL NETWORKS
CHAPITRE ONE
1.1.1 INTRODUCTION
H. Bhadeshia (2019). The basic definition of artificial neural networks is the modelling of the human brain, and the building blocks are neurons. The human brain contains approximately 100 billion neurons.
Each neuron has between 1,000 and 100,000 connections. knowledge is stored in the human brain in such a way that it can be spread, and we can pull more than one piece of this knowledge from our memory in parallel when necessary.
When we state that a human brain is made up of thousands of extremely strong parallel computers, we are not exaggerating. Neurons in multi-layer artificial neural networks are arranged similarly to those in the human brain.
Each neuron is linked to the others via specific coefficients. During training, information is delivered to these connection points in order to learn the network.the most advanced supercomputers.
Dr. Robert Hecht-Nielsen, the developer of one of the earliest neurocomputers, provides the simplest definition of a neural network, more appropriately known as a ‘artificial’ neural network (ANN).
A neural network, according to him, is a computing system composed of a number of simple, highly interconnected processing components that process information through their dynamic state reaction to external inputs.
ANNs are processing devices (algorithms or actual hardware) that are loosely modelled on much smaller sizes after the neuronal structure of the mammalian cerebral cortex.
A big ANN may have hundreds or thousands of processor units, but a mammalian brain has billions of neurons, with an increase in the magnitude of their overall interaction and emergent behaviour.
Although most ANN researchers aren’t interested with how closely their networks mirror biological systems, some are. For example, researchers have accurately mimicked retinal function and modelled the eye (Egmont-Petersen, 2020).
Artificial Neural Networks (ANNs)
1.2 Background Of Artificial Neural Network (ANN)
An Artificial Neural Network (ANN) is a data processing paradigm inspired by the way biological nerve systems process information, such as the brain. The unique structure of the information processing system is a crucial component of this paradigm.
It is made up of several highly interconnected processing components (neurones) that work together to address specific challenges. ANNs, like humans, learn by doing. Through a learning process, an ANN is trained for a specific application, such as pattern recognition or data classification.
In biological systems, learning entails changes to the synaptic connections that occur between neurons. This is also true for ANNs (Bishop and Christopher, 2015).
Artificial Neural Networks are made up of artificial neurons known as units. These units are grouped in a number of layers, which make up the entire Artificial Neural Networks of a system. The number of units in a layer can range from a few hundred to millions, depending on the system’s complexity.
In most cases, an Artificial Neural Network has an input layer, an output layer, and hidden layers. The input layer gets information from the outside world that the neural network must analyse or learn.
The data is then passed through one or more hidden layers, which change the input into data useful to the output layer. Finally, the output layer offers an output in the form of an Artificial Neural Network reaction to the input data (Borgelt and Christian, 2018).
The majority of neural networks connect units from one layer to the next. Each of these links has weights that define how much influence one unit has on another.
As data is sent from one unit to another, the neural network learns more about the data, resulting in an output from the output layer.
1.3 Advantages of Artificial Neural Network (ANN)
Storing information on the entire network: Information is stored on the entire network, rather than in a database, as in traditional programming. The absence of a few pieces of information at one location does not render the network inoperable.
Ability to work with incomplete knowledge: After ANN training, the data may provide output even when the information is partial. The loss of performance in this case is determined by the significance of the missing information.
Being tolerant of flaws: The corruption of one or more ANN cells does not preclude it from producing output. This feature makes networks more fault-tolerant.
Having a distributed memory: In order for an ANN to learn, examples must be determined and taught to the network according to the desired output by showing these instances to the network.
The network’s success is directly proportional to the number of instances selected, and if the event cannot be shown to the network in its entirety, the network may generate misleading information.
Gradual corruption occurs when a network slows and degrades over time. The network problem does not manifest itself quickly.
Machine learning capabilities: Artificial neural networks learn events and make decisions by commenting on similar events.
Parallel processing capability: Artificial neural networks have the numerical strength to perform several tasks at once.
1.4 Disadvantages Of Artificial Neural Network(Ann)
Hardware reliance: Because of their structure, artificial neural networks require processors with parallel processing power. As a result, the equipment’s realisation is dependant.
The most serious issue with ANN is unexplained network behaviour. When ANN generates a probing solution, it does not explain why or how. This undermines network trust.
The proper network structure is determined by no specific rule. Appropriate network structure is attained by trial and error.
Difficulty in communicating the problem to the network: ANNs can deal with numerical data. Before introducing problems to ANN, they must be converted into numerical values.
The display mechanism chosen here will have a direct impact on network performance. This is determined by the user’s skill.
The network’s lifespan is unknown: When the network’s error on the sample is lowered to a particular value, the training is complete. This value does not get the best outcomes.
Science artificial neural networks, which first appeared in the world in the mid-twentieth century, are fast evolving. We have now looked at the benefits of artificial neural networks as well as the issues that can arise when they are used.
It should not be forgotten that the problems of ANN networks, which are a developing science branch, are being eradicated one by one, while their benefits are expanding on a daily basis.
This suggests that artificial neural networks will become an increasingly significant element of our life (Cybenko, 2016).
Need help with a related project topic or New topic? Send Us Your Topic