Data fitting with neural networks

Data fitting is the process of building a curve or a mathematical function that has the best match with a set of previously collected points. Curve fitting can relate to both interpolations, where exact data points are required, or smoothing, where a flat function that approximates the data is built. We are talking about curves fitting in a regression analysis, which is most concerned with statistical inference problems, as well as the uncertainty that a curve coincides with observed data that has random errors. The approximate curves obtained from the data fitting can be used to help display data, to predict the values of a function where no data is available, and to summarize the relationship between two or more variables.

Figure 7.18: Linear interpolation of collected data

We have seen some examples of curve fitting in the regression...