|
|
|
|
|
|
 |
Introduction to Neural NetworksNeural Networks versus Conventional ComputingWhen considering the uses of artificial neural networks it is often useful to consider them in the context of conventional computing. Traditionally computers function as a "von-Neumann machine": instructions are fetched from memory, the CPU then processes data according to these instructions, and this process is repeated until all instructions are completed. This process is particularly well suited to solving problems using a series of well-defined steps (an algorithm) such as when searching for an item on a database. However, there are several limitations to conventional computing. First, the computer must be told in advance the details of the algorithm, often to great detail. However, even relatively simple tasks for people such as recognising faces are very difficult to express in a rigid algorithm. Second, even if an algorithm is known the data used must be as precise as possible. Conventional computers are often unable to manage the variability of data obtained in the real world. These two issues account for many of the areas that computers have traditionally been weak such as data prediction and classification. In contrast to von-Neumann machines, artificial neural networks - like our own brains - are well suited to situations that have no clear algorithmic solutions and are able to manage noisy imprecise data. This allows them to excel in those areas that conventional computers often find difficult. Index | Introduction | Neural Networks versus Conventional Computing © 2008 Marcus bros
|
|
|
|
