Keywords: gradient, SD method, CBB method, Minor Component
Abstract: We present a new gradient method for quadratic programming, named EM(Eliminating Minor Component Values), which eliminate the component in the direction of large eigenvalues when the current position is in the direction of small eigenvalues, thereby reducing the increase in the component of large eigenvalues in the next iteration. Numerical experiments show that the EM method has significant advantages over some commonly used optimization methods, such as BB and CBB.
Primary Area: optimization
Submission Number: 16768
Loading