What is a neural network?

What is a neural network?

© SolutionLibrary Inc. solutionlibary.com 9836dcf9d7 https://solutionlibrary.com/computer-science/theoretical-computer-science/what-is-a-neural-network-2gt

Solution Preview


Although this is a very simple example, it shows how the directed graph is able to calculate functions. More commonly, you will find large numbers of inputs. If you wanted to generalize the above example to n inputs, simply increase the number of input nodes to n, set all the weights w_1..w_n to 1, and have z calculate the product of all inputs. Again, a single input set to 0 will result in the output node returning 0.

Neural networks are rarely used in such a basic form, however. Most neural networks are layered, meaning that they consist of input and output nodes as above, but also add in one or more 'hidden' layers. Whereas the input nodes have weighted edges pointing at the output nodes in the above example, layered networks have input nodes that connect to hidden nodes, that in turn connect to output nodes. The following diagram shows one example:

In this example, we have 3 layers. Layer 0 is the input layer, on the far left. This is where the input values for the network arrive (for example, our boolean values for the AND-gate). Layer 1 is the hidden ...