Inductive Learning Algorithms
Hema R. Madala, Alexy G. Ivakhnenko · 2019
In problem solving, the main strategy is to specify a set of proper input-output associations and the main goal is to design an efficient learning algorithm. The network structures differ as per the interconnections among the units and their hierarchical levels. There are three main inductive learning networks: multilayer, combinatorial, and harmonic. The inductive learning algorithms can be divided into several main classes that could be constructed based on the addition (additive algorithms) or multiplication (multiplicative algorithms). In addition to these, there are other algorithms like correlational and orthogonalized (generalized) algorithms. The orthogonalized inductive algorithms allow one to improve stability in determination of coefficients. The principle characteristic of achieving an objective goal is for detailed (sharp) predictions in a low-level language which contain the greatest amount of detail while maintaining the prediction lead time that is typically obtained by using the most general high-level language.