Function-Based Classification and Regression Models Sample Clauses

Function-Based Classification and Regression Models. ‌ Function-based learning refers to the set of algorithms that solve or approximate a learning task via a mathematical function, as opposed to a logical function (as in tree-based methods). Whereas logical learning is usually associated with classification (but, as we have seen, is adapts perfectly to regression tasks), function-based learning is usually associated with regression, but all learned functions can als be used for classification. This is particularly true for modern function-based learning that is based on neural networks, but the argument works naturally with any learning function model. Linear regression is the most widely known function-based learning algorithm, and the literature on algorithms to approximate linear functions is wide and very well-known. Logistic regression is another widespread approach to function-based regression and, seemingly, can be seen as a particular case of neural networks or artificial neural networks. Artificial neural networks are an important tool that can adapt and adjust itself to the data, irrespective of any special functional or distributional requirement of the primary model. They are widely used for nonlinear pattern recognition and regression [75]. However, they are considered as black boxes due to the lack of transparency of internal workings and a lack of direct relevance of its structure to the problem being addressed, thereby making it difficult to gain insights. Furthermore, the structure of a neural network requires optimization which is still a challenge. Many existing structure optimization approaches require either extensive multi-stage pruning or setting subjective thresholds for pruning parameters. A Multilayer perceptron is a feed forward neural network that maps the inputs to an appropriate set of outputs and it is made up of several layers of nodes inside a graph, so that each layer is entirely linked to the next one with a nonlinear activation function, excluding the input nodes. Multilayer perceptron employs a supervised learning technique called back propagation for training purposes and a nonlinear activation function [76]. Support vector machines are another function-based technique that provides a classification learning model and an algorithm rather than a regression model and an algorithm [77]. This method is known as an algorithmic application of statistical learning theory and it uses some characteristics of the model and its performance on a training set to build reliable estim...