The conjugate gradient projection method is an important and efficient method to find the minimum value of the function for unconstrained optimization problems. In this work, we suggested a new three-term gradient projection technique with a nice descent property. In our technique, we mixed some conjugate gradient algorithms and projection techniques to solve a large scale of unconstrained optimization problems due to a limited memory. Under the Wolfe condition, we established the global convergent with some mild assumptions. The numerical results and performance profile indicate that the new formula is competitive.
Volume 12 | Issue 7
Pages: 788-795
DOI: 10.5373/JARDCS/V12I7/20202063