Traditional binary linear classification, maximal margin classification, support vector machines, and non-linear classification using kernel methods are covered in the chapter. Automated navigation on construction sites needs information about objects that are in the path of the vehicle or robot. Learning involves determining the values of weights and bias. Several algorithms have been developed for computing these values through iterative modifications using training data, one point at a time. The generalizability is poor in the perceptron learning algorithm because the algorithm terminates when it finds a hyperplane that does not cause any misclassification. The maximal margin classification can be solved using the dual formulation presented in the chapter. The optimization problem is solved in two stages. constraint solving; minimization of weights. In many applications, data are not linearly separable, that is, it is not possible to find a hyperplane that splits data into positive and negative classes.