Title :
A stochastic network with rotor neurons
Author :
Parra, Lucas ; Deco, Gustavo
Author_Institution :
Corp. Res. & Dev., Siemens AG, Munich, Germany
Abstract :
We define a new network structure to realize a continuous version of the Boltzmann machine (BM). Based on the mean field (MF) theory for continuous and multidimensional elements, named rotors (introduced by Gislen and Peterson), we derive the corresponding MF learning algorithm. Simulations demonstrate the learning capability of this network for continuous mappings.
Keywords :
Boltzmann machines; learning (artificial intelligence); neural nets; Boltzmann machine; continuous elements; continuous mappings; learning algorithm; mean field theory; multidimensional elements; network structure; rotor neurons; stochastic network; Annealing; Convergence; Equations; Multidimensional systems; Neural networks; Neurons; Research and development; Stochastic processes; Temperature; Tin;
Conference_Titel :
Neural Networks, 1993. IJCNN '93-Nagoya. Proceedings of 1993 International Joint Conference on
Print_ISBN :
0-7803-1421-2
DOI :
10.1109/IJCNN.1993.716805
