Roots of Nonlinear Functions



Illinois Algoritm

Illinois Algorithm

The Illinois algorithm is a variant of regula falsi which in turn is a variant of the secant method.  Given a continuous real-valued function of a real-variable, f(x), in which one wants to approximate the location of at least one root the Illinois algorithm proceeds after finding two values such that the function evaluated at the two values has opposite signs, i.e. if a and b are two numbers such that f(a)·f(b) < 0 then a zero of f(x) exists on the interval (min(a,b), max(a,b)).

The fact that a zero of f(x) exists in the interval (a,b) is a consequence of the intermediate value theorem of elementary analysis or from topology using the lemma that the continuous image of a connected set is connected as well as the lemma that a subset of R,(with the usual topology) with more than two points is connected if and only if it is an interval.

The Illinois algorithm iterative procedure uses two estimates as in regula falsi to estimate a new estimate, the estimate which is replaced is the one for which the sign of the function is the same as the sign of the function for the new estimate but if the same endpoint (right or left) is replaced in succession then the ordinate of the opposite endpoint is halved. The Illinois algorithm method is guaranteed to converge. The iterative process is continued until the halted. In the implementation given below the process is either terminated by the user, or when there is no machine representable number between the two endpoints, or when the function evaluates to zero at the new estimate.

Since the Illinois algorithm requires initial estimates in which the function evaluated at the initial estimates has opposite signs, it is not possible to use the Illinois algorithm to find the zeros of strictly nonnegative functions or nonpositive functions.

There are several comments which should be made in locating a zero of a continuous real-valued function of a real-variable f(x):
  • If α is any non-zero real number, then f(x) and α f(x) have the same roots. I.e. you can scale the function f(x) without modifying the location of the roots.

  • When the Illinois algorithm method terminates, the final interval is guaranteed to contain a zero of the function. But the final interval could also contain more than one zero of the function, in fact it could contain an uncountably infinite number of zeros.

Function List

  • int Illinois_Algorithm( double (*f)(double), double *a, double *fa,
    double *b, double *fb, void *data[],
    int (*verify_setup)(double, double, double, double, void**, int*),
    int (*terminate)(double, double, double, double, unsigned int, void**,int*),
    int *msg)

    Locate at least one root of f(x) in the interval ( min(*a,*b), max(*a, *b) ) where *fa = f(*a), *fb = f(*b) and *fa · *fb < 0. The procedure terminates when the user supplied routine terminate() returns a 1 in which case the Illinois_Algorithm returns a 2. Illinois_Algorithm returns a 1 if the user supplied routine verify_setup() returns a 0 which signifies that the initial setup has an error. If the user supplied routine verify_setup() returns a 1 or non-zero then the iteration process is started. Illinois_Algorithm can also terminate with different return values, see the table below.

    An example of a user supplied terminate() routine and a user supplied verify_setup() is given below in the routines Illinois_Algorithm_Terminate and Illinois_Algorithm_Verify_Setup respectively. Also included below is a routine to setup the array data called Illinois_Algorithm_Setup.

    The return values for the Illinois_Algorithm are:
    Return ValuesDescription
    0At least one of the pointer arguments is NULL.
    1The user supplied verify_setup() function returned a 1. The user specified error code should be stored at *msg.
    2The user supplied terminate() function returned a 1. The user specified termination code should be stored at *msg.
    3There is no machine representable number between the two endpoints.
    4The function evaluates to zero at the intersection of the secant (of the potentially modified function) and the x-axis.


    The arguments to Illinois_Algorithm are:
    ArgumentDescription
    fA real-valued function of a real-variable, i.e. a function with an argument of type double which returns a double.
    aA pointer to one end of the interval for which the function at the endpoints has opposite signs. When Illinois_Algorithm is invoked, *a is the initial value of an endpoint, on return *a is the final value of the endpoint.
    faA pointer to the value of f evaluated at *a, i.e. *fa = f(*a). When Illinois_Algorithm is invoked, *fa is the user supplied value of f(*a) where *a is the initial value at that endpoint, on return *fa is the final value of f(*a) where *a is the final value at that endpoint.
    bA pointer to the other end of the interval for which the function at the endpoints has opposite signs. When Illinois_Algorithm is invoked, *b is the initial value of an endpoint, on return *b is the final value of the endpoint.
    fbA pointer to the value of f evaluated at *b, i.e. *fb = f(*b). When Illinois_Algorithm is invoked, *fb is the user supplied value of f(*b) where *b is the initial value at that endpoint, on return *fb is the final value of f(*b) where *b is the final value at that endpoint.
    dataAn array of pointers of the type void*. The array data is not used by Illinois_Algorithm directly but only passed to the user supplied verify_setup() and terminate() routines. This allows the user to pass additional information to the terminate() routine so it can determine whether or not to terminate the iteration. The user supplied verify_setup() routine generally only checks that data has no obvious flaws.
    verify_setup()A user supplied routine to verify that the input data to Illinois_Algorithm and to the user supplied terminate() routine met the requirements that both routines function correctly. For a description of the arguments to verify_setup(), see the Illinois_Algorithm_Verify_Setup routine below which is an example of a user supplied verify_setup() routine. If verify_setup() detects an error, it can store an error code at *msg thus allowing the user to determine the reason for the error.
    terminate()A user supplied routine to terminate or continue iterating. For a description of the arguments to terminate(), see the Illinois_Algorithm_Terminate routine below which is an example of a user supplied terminate() routine. If terminate() returns 1 halt, it can store a halt code at *msg thus allowing the user to determine the reason for the halting.
    msgA pointer to an integer. This address is passed to the user supplied routines so that they can return error codes or halt codes.


  • int Illinois_Algorithm_Terminate( double a, double fa, double b, double fb, unsigned int iteration_count, void *data[], int *halt_code)

    Illinois_Algorithm_Terminate() is an example of a user supplied terminate() routine. This function returns a 0 if the process is to continue and a 1 if the iterating process is to halt. The arguments a, fa, b, fb of Illinois_Algorithm_Terminate() correspond to the dereferenced arguments a, fa, b, fb of the routine Illinois_Algorithm(), while data is the same, halt_code corresponds to argument msg of Illinois_Algorithm() and iteration_count is internal to Illinois_Algorithm().

    The return values for the Illinois_Algorithm_Terminate are:
    Return ValuesDescription
    0Continue iterating.
    1Terminate iterating.

    For this routine the data array consists of four pointers:
    Data Array IndexDescription
    0A (double *) pointer, cast as (void*), to the relative tolerance.
    1A (double *) pointer, cast as (void*), to the absolute tolerance.
    2A (double *) pointer, cast as (void*), to the lower bound of the region which determines whether to use the relative tolerance test or the absolute tolerance test to determine convergence.
    3A (double *) pointer, cast as (void*), to the upper bound of the region which determines whether to use the relative tolerance test or the absolute tolerance test to determine convergence.
    4An (unsigned int*) pointer, cast as (void*), to the maximum number of iterations allowed.

    If a 1 is returned (halt the iteration) then *halt_code is set to the following values:
    Halt CodeMeaning
    0The process converges using the relative tolerance test, i.e.
    |b - a| / min(|a|,|b|) ≤ relative_tolerance,
    where both a and b must have the same sign.
    1The process converges using the absolute tolerance test, i.e.
    |b - a| ≤ absolute_tolerance.
    2The function vanishes at one or both of the endpoints.
    3The number of iterations meets or exceeds the maximum allowable number of iterations without converging.


  • int Illinois_Algorithm_Verify_Setup( double a, double fa, double b, double fb, void *data[], int *err)

    Illinois_Algorithm_Verify_Setup() is an example of a user supplied verify_setup() routine. This function returns a 0 if an error was detected and a 1 if no error was detected, i.e. start the iteration process. The arguments a, fa, b, fb of Illinois_Algorithm_Verify_Setup() correspond to the dereferenced arguments a, fa, b, fb of the routine Illinois_Algorithm(), while data is the same and err corresponds to argument msg of Illinois_Algorithm().

    For this routine the data array consists of five pointers:
    IndexDescription
    0A (double *) pointer to the relative tolerance.
    1A (double *) pointer to the absolute tolerance.
    2A (double *) pointer to the lower bound of the region which determines whether to use the relative tolerance test or the absolute tolerance test to determine whether to halt the iteration or not.
    3A (double *) pointer to the upper bound of the region which determines whether to use the relative tolerance test or the absolute tolerance test to whether to halt the iteration or not.
    4An (unsigned int *) pointer to the maximum number of iterations allowed.

    If a 0 is returned (an error was detected) then *err is set to the following values:
    NumberMeaning
    1The endpoints agree i.e. a = b.
    2The function evaluated at the endpoints has the same sign. In case the function vanishes at an endpoint, the terminate() procedure will terminate the iteration.
    3The address of the data array is 0.
    4The address of an element of the data array is 0.
    5The relative tolerance is non-positive.
    6The absolute tolerance is non-positive.
    7The lower bound of the region which determines whether to use the relative tolerance test or the absolute tolerance test is not negative.
    8The upper bound of the region which determines whether to use the relative tolerance test or the absolute tolerance test is not positive.


  • int Illinois_Algorithm_Setup( double *rel_tolerance, double *abs_tolerance, double *test_region_lower_bound, double *test_region_upper_bound, unsigned int *max_number_of_iterations, void *data[])

    Illinois_Algorithm_Setup() is an example of a program to populate the array data for use in both Illinois_Algorithm_Verify_Setup() and Illinois_Algorithm_Terminate() routines after being passed to Illinois_Algorithm(). In particular,
    data[0] = (void *)rel_tolerance,
    data[1] = (void *)abs_tolerance,
    data[2] = (void *)test_region_lower_bound,
    data[3] = (void *)test_region_upper_bound, and
    data[4] = (void *)max_number_of_iterations,
    where *rel_tolerance is the relative tolerance, *abs_tolerance is the absolute tolerance, *test_region_lower_bound is the lower bound of the region which determines whether to use the relative tolerance test or the absolute tolerance test to determine whether to halt the iteration or not, *test_region_upper_bound is the lower bound of the region which determines whether to use the relative tolerance test or the absolute tolerance test to determine whether to halt the iteration or not and *max_number_of_iterations is the maximum number of iterations to try before terminating the iteration. The array data must be dimensioned at least 5 in the calling routine.

    The function, Illinois_Algorithm_Setup(), returns a 0 if at least one of the arguments is NULL, otherwise the function returns a 1.

C Source Code