Title :
Learning convergence in differential neurocontrol
Author :
Frolov, Alexander ; Roschin, Vadim ; Rizek, Stanislav
Author_Institution :
Inst. of Higher Nervous Activity & Neurophysiol., Acad. of Sci., Moscow, Russia
Abstract :
A new method of differential control is based on the analogy with the brain function in visual-motor coordination. The proposed scheme simulates both learning and operation procedures in living creatures. The method does not require calculation of matrix inverses. The model of pseudoinversion of Jacobean is generated as a layered neural network. It is being developed during the learning process based on the error backpropagation algorithm. Its convergence is analyzed in detail for a linear case
Keywords :
backpropagation; convergence; neurocontrollers; convergence; differential neurocontrol; error backpropagation algorithm; layered neural network; pseudoinversion; Artificial neural networks; Biological neural networks; Brain modeling; Computer science; Control systems; Convergence; Jacobian matrices; Neurocontrollers; Neurophysiology; Nonlinear control systems;
Conference_Titel :
Neuroinformatics and Neurocomputers, 1995., Second International Symposium on
Conference_Location :
Rostov on Don
Print_ISBN :
0-7803-2512-5
DOI :
10.1109/ISNINC.1995.480833
