Now that we have an idea of how to use regression to train a model, it’s time to explore the next step—fitting multiple regression units into a neural network.
Neural networks come as a better approach to solve complex regression or classification problems with non-linear dependencies between the input features. We have seen that linear regression and logistic regression with linear functions can perform well if the features of the underlying data are bound by linear relations. In that case, fitting a line (or a linear hyper-surface) through your data is a good strategy ...
