Representation before computation

My main objective is to point out a fundamental weakness in the conventional conception of computation and suggest a promising way out. This weakness is directly related to a gross underestimation of the role of object representation in a computational model, hence confining such models to an unrealistic (input) environment, which, in turn, lead to “unnatural” computational models. This lack of appreciation of the role of structural object representation has been inherited from logic and partly from mathematics, where, in the latter, the centuries-old tradition is to represent objects as unstructured “points”. I also discuss why the appropriate fundamental reorientation in the conception of computational models will bring the resulting study of computation closer to the “natural” computational constrains. An example of the pertinent, classoriented, representational formalism developed by our group over many years—Evolving Transformation System (ETS)—is briefly outline...

• Type: Journal
• Year: 2010
• Where: NC
• Authors: Lev Goldfarb
• Keywords: Appropriate Fundamental Reorientation, Computational Models, NC 2010, Neural Networks, Object Representation