The talk presents an overview of current methods of computational intelligence (CI) called evolving CI (eCI) and how they can be used in to create adaptive, computational intelligence (CI) systems across areas of applications. Evolving systems evolve their structure and functionality in a self-organised, adaptive, incremental way to capture patterns form input data. The methods presented include: evolving connections systems (ECOS) and evolving neuro-fuzzy systems in particular; evolving spiking neural networks (eSNN); evolutionary and neurogenetic systems; quantum inspired evolutionary computation; rule extraction from ECOS and eSNN. The methods above are suitable for incremental adaptive, on-line learning from data and data mining. They are applied on spatio and spectro temporal data modeling and pattern recognition problems, including: moving objectrecognition, gesture- and sign language recognition; bioinformatics; ecological and environmental modeling, such as establishment and spread of invasive species; cybersecurity; brain data modeling and brain-computer interfaces.eSNN have? proved superior for spatio and spectro-temporal data analysis, modeling and pattern recognition (http://ncs.ethz.ch/projects/evospike/). Future directions for eCI are discussed including hardware-software system development and neuromorphic engineering. Materials related to the lecture, such as papers, data and software systems can be found on the Knowledge Engineering and Discovery Research Institute KEDRI web site (www.kedri.info) of the Auckland University of Technology and the Rio de Janeiro Brazilian Chapter of the IEEE CIS.
展开▼
