The basics of multi-layer feedforward neural networks / Nurul Aityqah Yaccob and Farizuwana Akma Zulkifle

Artificial neural networks are computational models inspired by the brain, enabling them to capture complex nonlinear relationships between a response variable and its predictors. The simplest networks lack hidden layers, making them equivalent to linear regression models. Figure 1 illustrates a neural network representation of a linear regression model with four predictors. The coefficients assigned to these predictors are called "weights," and the forecasts are generated through a linear combination of the inputs. The weights are selected in the neural network framework using a "learning algorithm" that minimizes a "cost function," such as the mean squared error (MSE). However, for this simple case, 'near regression remains a more efficient approach.