Onedimensional minimization

Please choose the method:
Golden section search
Iterated quadratic interpolation
Iterated backtracking Armijo-parameter δ :
δ = Important: 0 < δ < 0.5
Safety constants c1 and c2 for choosing the initial stepsize by interpolation, must be in [c1,c2]:
c1 = Important: 0 < c1 << 1
c2 = Important: 1 << c2
Iterated descent using the Powell-Wolfe stepsize
( Albaali and Fletcher method)
Parameter for Powell-Wolfe:
δ = Important:0 < δ <0.5
κ = Important :0 < δ < κ

Please choose a function:
f(x) = exp(-2*x)+3*exp(x), uniformly convex
f(x) = -(1/(x2+1)+1/((x-1)2+1)), quasiconvex
f(x) = 0.25*x4+2*x3+4.5*x2-16*x , quasiconvex
f(x) = 2*exp(-0.1*x)+exp(10*x), uniformly convex, but badly conditioned
a function of your own:
Please type the evaluation program of your function here using FORTRAN rules. Your final statement must be
      fu= some expression you computed before or just here depending on x
You may use the constants pi, e(=exp(1)), sqrt2(=1.414...), the integer variables i,j,k, the logicals bool1,bool2,bool3 and the double precision variables sum,h1,h2,h3,h4,y(100),z(100),a(100,100) which are all intialized with zero resp. .false. . The routine has the parameters x (double, input) and fu (double out). never change x!. first is a local integer and set 0 before calling the function the first time. You may use this in order to initialize some local data and set it 1 afterwards to avoid multiple such initialization. Your settings of the local variables are preserved during program execution.


Please specify the interval here:Important: a < b and f'(a)*f'(b) < 0
a = b =

Please specify the desired precision
eps =

Click on "evaluate", in order to submit your input.

back to the theory page

 Back to the top!

16.05.2016