I am developing a code that uses Newton's method to optimize a functional W:ℝ6→ℝ Because the functional is poorly represented by a quadratic, convergence is slow. Therefore I've implemented a line search using two methods: the secant method and the backtracking method (with the Armijo condition) My problem is that the line search produces a factor of zero. When plotting W(x + α Δx) [itex] \ \ \ \ \ [/itex] α[itex]\in[/itex][-1,1] as a function of [itex]\alpha[/itex], it is clear that α=0 is a minimizer. I can't understand why this would be the case. Thanks in advance.