gradient-descent-armijo-wolfe

Bespoke, from scratch, implementation of Armijo-Wolfe inexact line search technique to find step length for gradient descent optimisation. The library alternative is scipy.optimize.line_search

Example inputs provided to program:

1. roll = 'M21AI619'
2. R = last 2 digit of roll dynamically obtained
3. FX = (x1 − R)^2 + (x2 − 2x1)^2
4. delFX = dynamically calculated from FX
5. epsilon = 10^-3
6. alpha_1 = 1.0
7. r = 0.5
8. beta_1 = 10^-4
9. beta_2 = 0.9
10. X0 = [R+3,2R-2]
11. D = dynamically calculated from delFX
12. k = 1

Example output of program:

1. F(x+alpha*dK): [18.99935447 37.99813236]
2. Alpha (step length): 0.125
3. Dk (descent direction): [-0.00101529 0.00115317]
4. K (number of iterations): 323