Neural Networks| Neural Networks |
|
Artificial Neural Networks attempt to simulate in a mathematical manner the human neural system and thus, mimic the distributed operation of the brain. They are consisted of simple structural elements called neurons, which can be controlled by certain parameters and are able to "learn" and respond in a "smart" way to any new stimulus. Neural networks' learning ability follows the same procedures as those experienced by a human being during the first steps of his or her life. Consider how a little child finds out that the kitchen's ceramic hot-plate is dangerous when it is red and radiates heat, while it is safe when it is black and cold. Neural networks train their "brain" in a similar way, in order to classify situations or objects into categories based on patterns. They perform a process of computational reasoning which combines the current characteristics of the problem at hand and previously aqcuired knowledge concerning it. Usually, the knowledge of the past is given to them in the form of examples, much like a teacher instructs its pupils. Neural networks, today, are able to address not only pattern classification tasks, but also tasks concerning control, clustering, regression or time series prediction. |