Artificial neural networks (ANNs) were built to model the brain for the purpose of solving the problems humans alone cannot as well as to advance, artificial intelligence. To approximate organic beings and gain great computational power, to become a technological hybrid between sentient beings and advanced electronics; they are the future of advanced robotics.
They can be used in miscellaneous fields such as speech recognition, prediction of stocks, weather and so on.
Artificial neural networks (ANNs) approximates the probable function that will likely produce the best output. This is done through extensive training of the system and the use of ample training rules which allows the ANNS to recognise repetitive paradigms and use them to solve problems that the system has not encountered before. These systems model the mammalian cerebral cortex (the brain) and its neurons, hence the name artificial neural networks. Before understanding the complex structure of an artificial neural network, a rudimentary knowledge of an organic neural network is essential.
The human brain consists of over billions of neurons interconnected by trillions of synapses. Neurons exchange electrical impulses through the synapses attained from other neurons or from the senses. When something novel is experienced, these neurons create new connections which may weaken, fortify or alter through time. These are the experiences or the memories that humans recollect and is a basis for the fundamental of decision making and problem solving. Artificial neural networks use these same principles; they model an approximate function based on the input and output rules and use this function to predict the output for a problem that the system has never faced. The simplest example is an ANN learning the XOR function (Exclusive OR) which has two binary input variables and a binary output variable with only four possible training rules.
If the two inputs (input A and Input B) are alike then the output is 0 and if the inputs are different then the output is 1, as shown by the table above. The use of this rudimentary function will mean that the ANN will only see 4 possible inputs which are also the rules, however more advanced ANNs will encounter inputs that are not part of the input rules.
The above XOR function is a very basic example used solely for understanding the purpose of an ANN. Real life examples are much more complex with many input variables, countless training rules and multiple outputs and does not guarantee 100 percent accuracy. This concept of accuracy will be discussed further in the report.
A neural network is created by modelling the natural neurons and creating mathematical artificial neurons. An ANN is constructed of multiple layers of neurons. The first layer consists of input neurons while the last layer consists of output neurons. In the middle are some hidden layers which differ, upon individual ANNS. All neurons in one layer are connected to the next layer by...