Abstract
After a short introduction to neural networks generally, a more detailed presentation of the structure of a feed forward neural network is done, using mathematical language, functions, matrices and vectors.
Further, emphasis has been made on perceptrons and linear regression done by using ANN. Central concepts like learning, including weight updates, error minimization with gradient descent are introduced and studied using these simple networks. Finally, multilayer perceptrons are defined with their error functions and finally backpropagation are described precisely using composite functions and the concept of error signals.Keywords: Backpropagation, Chain rule, Composite functions, Computing neurons, Feedforward, Matrices, Multiple perceptron, Neural network, Perceptron, Transfer function, Vectors.
About this chapter
Cite this chapter as:
Hans Birger Drange ;A Mathematical Description of Artificial Neural Networks, Artificial Intelligence: Models, Algorithms and Applications (2021) 1: 117. https://doi.org/10.2174/9781681088266121010010
DOI https://doi.org/10.2174/9781681088266121010010 |
Publisher Name Bentham Science Publisher |