2
$\begingroup$

I'm in a math workshop, where one of the problems given was $y=–3x^{1.25} –3x+10$. Much to my frustration, the only stated way to find roots was finding x by trial and error.

Is there any way to accurately find the roots of this function using a root-finding algorithm – not unlike the quadratic formula?

$\endgroup$
  • $\begingroup$ There is nothing to solve here. You mean $y=0?$ $\endgroup$ – L. F. Aug 19 '13 at 17:00
  • $\begingroup$ Yes, I meant finding the function's roots, which I thought was synonymous. $\endgroup$ – Jules Aug 19 '13 at 17:04
  • $\begingroup$ That is precisely what @L.F. meant. $\endgroup$ – Ron Gordon Aug 19 '13 at 17:05
  • 2
    $\begingroup$ @atleastthreecharacters No, they are not synonyms. The term solving makes sense when talking about equations, not when talking about functions. $\endgroup$ – Git Gud Aug 19 '13 at 17:08
  • $\begingroup$ @GitGud thanks for the correction. Fixed the question. $\endgroup$ – Jules Aug 19 '13 at 17:09
5
$\begingroup$

Write $x=u^4$. Then we have to solve $$ -3u^5-3u^4+10=0$$

A numerical solution $u\approx1.1198125298329972524296485747877379384$ (so $x\approx 1.5724660961391427815788005324759241792$) is readily found, however an exact solution with radicals is ... unlikely. At least for general equations of fifth degree there is (provably!) no such solution method possible. One would need to go into some deeper detail to check if this specific equation also has no solution by radicals. Actually doing this would be overkill for a precalculus course and Thus I won't, but I bet it is the case.

$\endgroup$
  • $\begingroup$ Excellent idea! Thanks! $\endgroup$ – Jules Aug 19 '13 at 17:13
2
$\begingroup$

There are algorithms for solving this, but they require calculus. One of the most commonly used methods is Newton's Method, which approaches the root using tangent lines. Let $y=f(x)$ and choose a seed $x_0$. To get $x_{n+1}$ from $x_n$, first compute the tangent line: $$\tilde{y}_n=f(x_n)+f'(x_n)(x-x_n),$$ where $f'$ represents the first derivative of $f$ with respect to $x$. From there, find the $x$-intercept by setting $\tilde{y}_n=0$ and solving for $x$. This intercept is your $x_{n+1}$. In short, $$x_{n+1}=x_n-\frac{f(x_n)}{f'(x_n)}.$$ This iterative method converges to the root of the function. For a faster convergence, you can add additional terms from the Taylor polynomial $$f(x)=\sum_{k=0}^\infty \frac{f^{(k)}(x_n)(x-x_n)^k}{k!}.$$

$\endgroup$
1
$\begingroup$

I assume that the question is to find the roots of the equation you've given. That is, we need to find $x$ such that $$ 3x^{1.25} + 3x - 10 = 0 $$ We can rearrange the question to get $$ 3x^{1.25} = 10 - 3x\\ (3x^{1.25})^4 = (10 - 3x)^4\\ 81x^5=81 x^4-1080 x^3+5400 x^2-12000 x+10000\\ 81x^5-81 x^4+1080 x^3 - 5400 x^2 + 12000 x - 10000=0 $$ The question then becomes "what $x$ is the real root of the above equation?" As it ends up, there is no neat solution like the quadratic formula for polynomials of degree $5$ or higher, and this seems to be an example of an irreducible quintic equation. Trial and error does seem to be the best approach after all.

$\endgroup$
  • $\begingroup$ Well one can estimate the solution graphically and then use e.g. binary search, this isn't quite 'trial and error'. $\endgroup$ – Ruslan Aug 19 '13 at 17:11
  • $\begingroup$ @Ruslan I would tend to think of binary search and the newton raphson methods as a particular type of trial and error, though perhaps my notion differs from the accepted definition $\endgroup$ – Omnomnomnom Aug 19 '13 at 17:16
  • $\begingroup$ @Ruslan Better yet, Newton's Method. $\endgroup$ – Gamma Function Aug 19 '13 at 17:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.