Abstract
Different neural net node computational functions are compared using feedforward backpropagation networks. Three node types are examined: the standard model, ellipsoidal nodes and quadratic nodes. After preliminary experiments on simple small problems, in which quadratic nodes performed very well, networks of differing nodes types are applied to the speech recognition 104 speaker E-task using a fixed architecture. Ellipsoidal nodes were found to work well, but not as well as the standard model. Quadratic nodes did not perform well on the larger task. To facilitate an architecture independent comparison a transputer-based genetic algorithm is then used to compare ellipsoidal and mixed ellipsoidal and standard networks with the standard model. These experiments confirmed the earlier conclusion that ellipsoidal networks could not compare favourably with the standard model on the 104 speaker E-task. In an evolutionary search in which layer node types were free to adjust ellipsoidal nodes had a tendency to become extinct or barely survive. If the presence of ellipsoidal nodes was enforced then the resulting networks again performed poorly when compared with the standard model.
Similar content being viewed by others
References
Jones AJ. The modular construction of dynamic nets. Neural Computing & Applications 1993; 1(1)
Shepard RN. Stimulus and response generalisation: Deduction of the generalisation gradient from a trace model. Psychological Review 1958; 65: 242–256
Shepard RN. Towards a universal law of generalisation for psychological science. Science 1987; 237
Tawel R. Does the neuron ‘learn’ like the synapse? In: Advances in Neural Information Processing 1, ed. D.S. Touretzky, San Mato, CA: Morgan-Kaufmann, 1989; 169–176
Jones AJ. Genetic algorithms and their applications to the design of neural networks. Neural Computing & Applications 1993; 1(1)
Bimbot F. Speech processing and recognition using integrated neurocomputing techniques. Technical report BRA3228, Cap Gemini Innovation, 118 rue de Tocqueville, 75017 Paris, France
Dodd N, Macfarlane D, and Marland C. Optimisation of artificial network structure using genetic techniques implemented on multiple transputers. In: Transputing 91, Vol 2. Eds D Styles, T Kunii, A Bakkers, IOS Press, 1991
Macfarlane D, East I. An investigation of several parallel genetic algorithms. OUG-12: Tools and Techniques for Transputer Applications. IOS Press, 1990.
Giles C, Griffin R, Maxwell T. Encoding geometric invariances in higher order neural networks. Neural Information Processing Systems, American Institute of Physics, New York, 1988; 301–309
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Jones, A.J., Macfarlane, D. Comparing networks with differing neural-node functions using transputer-based genetic algorithms. Neural Comput & Applic 1, 256–267 (1993). https://doi.org/10.1007/BF02098744
Received:
Revised:
Issue Date:
DOI: https://doi.org/10.1007/BF02098744