A Kernel - Based Model for Artificial Immune Networks

An artificial immune network model is a bioinspired technique to machine learning, which is defined upon some immune concepts and principles, mainly the so-called idiotypic network theory. This kind of models are mostly feature-based techniques, that is, they assume a vector representation of the input data. Similarity-based learning techniques do not assume a particular representation of the input data. Instead of it, they assume the existence of a similarity measure defined on the input space. Kernel methods allow to define similarity-based learning techniques by decoupling data representation from the algorithm dynamics. This work presents a strategy to build artificial immune network models following a similarity-based approach through the use of concepts from kernel methods.