The paper explains the principal concepts behind the conventional multiplayer feed forward network and defines the terminology that is required to understand the literature in this area. It starts by explaining the structure and functionality of the prototype feed forward network. After outlining some of its recent applications, its connections with biological and statistical models is briefly discussed. The answers to the questions what is this network capable of? and 'how difficult is it to train?' are the topic of the next two sections. Pre-training data processing techniques and an overview of methods for network training are then presented. The ability of a network to learn the main concepts hidden in noisy data are also discussed. Some examples of its implementation in hardware implementation requirements, the capabilities of networks with constraints placed on their weights are then explained. The paper concludes with a short note on the major milestones in the development of feedforward networks.