Nature-Inspired Learning Models.

Abstract

Intelligent learning mechanisms found in natural world are still unsurpassed in their learning performance and eficiency of dealing with uncertain information coming in a variety of forms, yet remain under continuous challenge from human driven artificial intelligence methods. This work intends to demonstrate how the phenomena observed in physical world can be directly used to guide artificial learning models. An inspiration for the new learning methods has been found in the mechanics of physical fields found in both micro and macro scale. Exploiting the analogies between data and particles subjected to gravity, electrostatic and gas particle fields, new algorithms have been developed and applied to classification and clustering while the properties of the field further reused in regression and visualisation of classification and classifier fusion. The paper covers extensive pictorial examples and visual interpretations of the presented techniques along with some testing over the well-known real and artificial datasets, compared when possible to the traditional methods.