Biologically-Inspired Adaptive Learning: A Near Set Approach

James F. Peters, Shabnam Shahfar, Sheela Ramanna, Tony Szturm
2007 2007 Frontiers in the Convergence of Bioscience and Information Technologies  
The problem considered in this paper is how learning by machines can be influenced beneficially by various forms of learning by biological organisms. The solution to this problem is partially solved by considering considering a model of perception that is at the level of classes in a partition defined by a particular equivalence relation in an approximation space. This form of perception provides a basis for adaptive learning that has surprising acuity. Viewing approximation spaces as the
more » ... counterpart of perception was suggested by Ewa Orłowska in 1982. This view of perception grew out the discovery of rough sets by Zdzisław Pawlak during the early 1980s. The particular model of perception that underlies biologically-inspired learning is based on a near set approach, which considers classes of organisms with similar behaviours. In this paper, the focus is on learning by tropical fish called glowlight tetra (Hemigarmmus erythrozonus). Ethology (study of the comparative behaviour of organisms), in particular, provides a basis for the design of an artificial ecosystem useful in simulating the behaviour of fish. The contribution of this paper is a complete framework for an ethology-based study of adaptive learning defined in the context of nearness approximation spaces. Index Terms- Approximate adaptive learning, behaviour, ethology, machine learning, near set, observation, perception. An approximation space ... serves as a formal counterpart of perception ability or observation. , set of partitions, ν Nr ν Nr : P(O) × P(O) −→ [0, 1], overlap function, Nr(B) * X Ë x:[x] Br ⊆X [x] Br , lower approximation, Nr(B) * X Ë x:[x] Br ∩X [x] Br = ∅, upper approximation, Bnd Nr (B) (X) Nr(B) * X\Nr(B) * X.
doi:10.1109/fbit.2007.39 dblp:conf/fbit/PetersSRS07 fatcat:lcsqdogjbje75mq2b775l5ecg4