Neural implementation of Karmarkar algorithm
P. Sridhar, Ujjwal Kumar, Arun K. Pujari · 1991
The Karmarkar algorithm performs a sequence of projective transformations each followed by optimization over an inscribed sphere and then inverse projective transformation. These steps are implemented on the authors' model with two neural-like circuits, one embedded inside the other. The inner circuit finds least square error solutions using the complementary slackness condition. The outer circuit finds a novel interior primal solution using the delta rule by making use of the error in the computation of dual solution. This circuit exhibits potential for applications where real-time optimization is required-as is the case in robotics, satellite guidance, etc.>