Python gives me different outputs if I add a print() later on

Question

I'm super new to programming and just wanted to make something to find the approximate value of x for the equation x^3 - 1 = x . I know that -2 is greater than 0 and -1 is less. My thought is that if I find the average and check if it's greater or less than 0 , I can redefine a and b and repeat until I get an approximate value. I've been having a hard time getting it to act right though. For example if I run this block without the print(i), I'll get -1.5 which would be the average, but when I put a print(total) within the function equation(n) to see if it's working right, it doesn't even show me that and just outputs -8.881784197001252e-16 . If I put print(i) at the end of the if/else possibilities, such as this, it prints 16 and then -8.881784197001252e-16 . I'm using PyCharm CE.

Beyond this glitch, is my logic correct? By setting the placeholder to 1 it will run the while loop. The while loop will get a new value of n and run the function, compare it to 0 , then reassign a or b depending on that comparison? Thanks in advance.

a = float(-2)
b = float(-1)
n = ((a+b)/2)
print(n)

def equation(n):

    total = float((n - n**3 - 1))

    return total

i = 1

while i != 0:
    n = ((a + b) / 2)
    if (equation(n)) > 0.0:
        a = n
        i = equation(n)
        print(i)

    else:
        b = n
        i = equation(n)
        print(i)

Answer 1

Moving all of the equation x^3 - 1 = x to one side should give x^3 - 1 - x = 0 or x - x^3 + 1 = 0 . You have a different equation in your function.

Another problem is that there is no intersection between the two equations between x=-2 and x=-1 (see here ). You'll need to expand your window to x=2 before you'll see an intersection.

Something that often happens in numerical analysis (where you'll see this type of problem) is that rather than trying to find x that actually makes the equation give 0 , we look for a value of x that produces below an acceptable level of error. Another approach is to test for while b - a > tol:

If we use all of this to tweak what you've got, you'll have

a = float(-2)
b = float(2)
tol = 0.001

def equation(n):
    return float(n - n**3 + 1)

n = (a + b) / 2
iter = 0
while abs(equation(n) - 0) > tol and iter < 100:
    iter+=1

    if equation(n) > 0.0:
        a = n
    else:
        b = n

    n = (a + b) / 2

    print(iter,a,b,equation(n))

---

Note: this works fine if you remove the floats and just do

a = -2
b = 2
#...etc

because python already recasts values as necessary. Try

>>> type(3)
<class 'int'>
>>> type(3.5)
<class 'float'>
>>> type(float(3))
<class 'float'>
>>> type(3/5)
<class 'float'>

so python will store the result as a float as soon as it is necessary.

Answer 2

The immediate problem is the limited precision of floats. If you print a and b after a hundred iterations, you get:

a, b = -1.324717957244746, -1.3247179572447458
print((a + b) / 2  # -1.3247179572447458, the same as b

So at some point, b never changes, which is why you get an infinite loop. If we evaluate equation at the average of a and b , you'll get -8.881784197001252e-16, the value you were always seeing.

But this will never converge to exactly zero because the solution is irrational , so even if you had infinite precision, equation would never equal zero.

The common way to resolve this issue is to avoid comparing floats:

if a == b               # don't do this
if abs(a - b) < epsilon      # do this, where epsilon is some small value

(Note: what you're describing is the bisection method , which is slower than higher order algorithms eg Newton's method, which you can use since you can get the derivative)