4 Artificial Neural Networks
The study of artificial neural networks help us understand the brain, perception and human cognition and lead to the development of systems with value-added functionality to solve difficult problems. This means the development of neural networks is inspired by the human capabilities for learning, understanding and performing complex tasks with easiness and accuracy. To reference something, simply recognizing human faces, kicking a football or fetching a glass of water, things we do without even realizing or knowing how these tasks are achieved in our brain. Developing a technology with abilities to efficiently perform the same tasks as humans, but more faster, accurately and without ever getting tired or minding working in unsustainable conditions has been a driving force for researchers in academies and in the industry. Neural networks has been proved that they can efficiently solve many problems that are difficult using conventional programming techniques. Successful approaches have been achieved in prediction, pattern recognition, categorization and approximation fields.
4.1 Introduction to neural networks
4.1.1 The biological neural networks
It is well-known that there exist over hundred categories of neurons; however most of their common features have been found. The biological neural network comprises of large number of nerve cells (neurons) which are linked with each other via synapses. The neurons number is determined at the birth and is approximately estimated to 10 billion and is fairly constant throughout the life. New born human has less synapses compared to an adult, but develops hundreds of thousands new ones per second in response to the learning of the new world.
Figure 4.1 A ...
... middle of paper ...
...PROP on the other hand only sees to the sign of the gradient and discards big magnitudes. This means it is only important if the gradient is positive, negative or near zero. If the magnitude is small the RPROP won’t change any weight matrix. If the magnitude is positive the weight will be increased by a constant, if it is negative it will be decreased by this constant. This constant has to be provided to the algorithm.
The RPROP uses special so called “deltas” that it calculates by itself, these “deltas” are not global, and there is one for each threshold and weight in the matrix. In the beginning each of these “deltas” is set to a very small number, and as the training succeeds they will increase or decrease with the help of the error gradients sign. By doing so each weight matrix will be individually adjusted and this is a huge advantage over the backpropagation.
Kandel, E. R., J. H. Schwarz, and T. M. Jessel. Principles of Neural Science. 3rd ed. Elsevier. New York: 1991.
The human brain and that of any species contains nerve cells that link to each other connecting the brain and spinal cord to the rest of the body (Johnson, 2013). These nerve cells are neurons that connected through synapses in a web-like fashion forming neural networks (Coon & Mitterer, 2001). Neural networks make generation and transmission of action potentials (known as electrical impulses) possible along neurons. An action potential is generated across an axon hillock of a nerve cell and is propagated along the axon by the opening of voltage-gated ion channels one after the other causing positive ions to flow in and out the axon (Johnson, 2013).
Stergiou, C., & Siganos, D. (2011, August 6). Neural Networks. Retrieved August 6, 2011, from
According to Berlucci and Butchel (2009), plasticity describes an alteration in neural organization. Plasticity may be to blame for several types of behavior changes both short-term and permanent, such as growth, learning, injury, aging, and adaption to various settings. While several authors have attempted to more appropriately define the term, researchers are inclined to relate the theory to essentially any deviation found within the nervous system. Today, the method of behavior change can mostly likely be described by the alteration of synaptic transmission amongst neurons.
Brains and the more nonessential parts of the nervous system are made up of neurons (Meadows, 1993 p. 266). The brain is made up of neurons, axons, dendrites, and synapse. They all have their individual job that contributes to the functioning of the body. The weight of the brain increases as a child develops into adulthood. According to Siegler & Alibali (2005, p13), the changes in the size of the brain make it possible for advanced thinking. The neurons are the functioning core of the brain (Brotherson, 2005). The brain begins to develop in the mother’s womb and continue to develop as the child develops. The neuron has branches protruding from the cell sending signals to the synapse and axon. The synapse and axon shapes the brain which allows connections to be made. Young children learn new information when they follow the same routine on a regular base. If a parent repeatedly calls a child a certain name, then connections form that allow the child to recognize that name over time and he or she will begin to respond to that name (Brotherson, 2005). Through repetitive experiences the axons and synapse strengthen causing learning to take place.
"My name is Dorothy," said the girl, "and I am going to the Emerald City, to ask the Oz to send me back to Kansas."
The number of synaptic inputs recieved by each nerve cell in our (human) nervous system varies from 1-100,000! This wide range reflects the fundamental purpose of nerve cells, to integrate info from other neurons.
As our research into science and technology ever increases its seems inevitable that in the near future Artificial Intelligent machines will exist and become part of our everyday life such as we see with modern computers today.
Biology The brain consists of both neurons and glia cells. The neurons, which are cells housed in a cell body called a Soma, have branches which extend from them, referred to as dendrites. From these dendrites extend axons which send and receive impulses, ending at junction points called synapses. It is at these synapse points that the transfer of information takes place. At the heart of neuroplasticity is the idea of synaptic pruning.
The Central Nervous System (CNS) consists of approximately 100 billion linked neurons, furthermore, it can be categorised into two major divisio...
The approach to artificial intelligence should be proceeded with caution. Throughout recent years and even decades before, it has been a technological dream to produce artificial intelligence. From movies, pop culture, and recent technological advancements, there is an obsession with robotics and their ability to perform actions that require human intelligence. Artificial intelligence has become a real and approachable realization today, but should be approached with care and diligence. Humans can create advanced artificial intelligence but should not because of the harm they may cause, the monumental advancement needed in the technology, and that its harm outweighs its benefits.
The most basic elements of a neural network, the artificial neurons, are modeled after the neurons of the brain. The "real" neuron is composed of four parts: the dendrites, soma, axon, and the synapse. The dendrites receive input from other neuron's synapses, the soma processes the information received, the axon carries the action potential which fires the neuron when a threshold is breached, and the synapse is where the neuron sends its output, which are in the form of neurotransmitters, to the dendrites of other neurons. Each neuron in the human brain can connect with up to 200,000 other neurons. The power and processing of the human brain comes from multitude of these basic components and the many thousands of connections between them.
Artificial Intelligence is the scientific theory to advance the scientific understanding of the mechanisms underlying thought and intelligent behavior and their embodiment in machines. This is going to hold the key in the future. It has always fa...
Singh, Y. & Chauhan, A. (2009). Neural networks in data mining. Journal of Theoretical and
Machine learning systems can be categorized according to many different criteria. We will discuss three criteria: Classification on the basis of the underlying learning strategies used, Classification on the basis of the representation of knowledge or skill acquired by the learner and Classification in terms of the application domain of the performance system for which knowledge is acquired.