Читать книгу Digital Forensics and Internet of Things - Группа авторов - Страница 17

A neural network is an instrument that is designed to model in the similar way in which the brain responds or executes a task or function; it is usually simulated in digital computer-based software or carried out by using electronic components. It can resemble the brain in the following aspects:

 • The knowledge is obtained by the network from its surrounding with the help of a learning procedure.

 • Interneuron link strength, known as synaptic weight, is used to accumulate the obtained knowledge.

 • The process that is operated to execute the learning procedure is known as the learning algorithm; the purpose of which is to reform the synaptic weights of the network in a well-organized mode to accomplish the desired layout objective.

 • It is also possible to improve its own topology.

 • Neural network is also mentioned in literature as neurocomputers, connectionist network, and parallel distributed processor.

 • Neural network attains its computing power at the beginning from its power of computer at first from the massively side-by-side distributed arrangement and next from its potential to learn and then generalize.

 • Generalization leads to the neural network constructing logical outputs for inputs not encountered throughout training (learning).

An ANN is specified by the following:

 • Neuron model: Data processing component of the neural network.

 • An architecture: A group of neurons along with connections connecting neurons.

 • A training algorithm: It is used for instructing the Neural network by changing the weights to model a selected training task correctly on the instructing examples.