Fibonacci challenge

gudii9
gudii9 used Ask the Experts™
on
Hi,

I am working on below challenge
http://codingbat.com/prob/p120015

Psedo code:
1. if first two numbers return same
2. else call same fibonacci method by passing n-1 and n-2 and sum it up

public int fibonacci(int n) {
  
   if ((n == 0) || (n == 1)) // base cases
      return n;
    else
      // recursion step
      return fibonacci(n - 1) + fibonacci(n - 2);
  

}

Open in new window




I am passing all tests
Expected      Run            
fibonacci(0) → 0      0      OK      
fibonacci(1) → 1      1      OK      
fibonacci(2) → 1      1      OK      
fibonacci(3) → 2      2      OK      
fibonacci(4) → 3      3      OK      
fibonacci(5) → 5      5      OK      
fibonacci(6) → 8      8      OK      
fibonacci(7) → 13      13      OK      
other tests
OK      

How to improve/modify my design, code and any other alternate approaches. please advise
Comment
Watch Question

Do more with

Expert Office
EXPERT OFFICE® is a registered trademark of EXPERTS EXCHANGE®
mccarlIT Business Systems Analyst / Software Developer
Top Expert 2015
Commented:
No improvement necessary, that's about as simple as it gets!
Top Expert 2015
Commented:
You could get rid of the extra brackets
ste5anSenior Developer
Commented:
This again: These tests and tasks are pretty poor. Fibonacci is definied here over N0. Thus a test covering the input -1 should exist. And you should test the input first.
Announcing the Winners!

The results are in for the 15th Annual Expert Awards! Congratulations to the winners, and thank you to everyone who participated in the nominations. We are so grateful for the valuable contributions experts make on a daily basis. Click to read more about this year’s recipients!

No changes needed - you've nailed this one.

Doug
I agree with ste5an. On the other hand, this is for learning the basics of programming, and the totally correct solution (with checking of everything, with unit test, and so on) is not necessary to understand the principle (of Fibonacci, and of the recursive solution).

What is important in this case is that you should not remember this recursive solution as the right approach. The reason is that it is extremely ineficient. The non-recursive solution is not more difficult here, and it is much more efficient.
More advanced ways to compute fibonacci(n) really fast - no recursion, and no loops to get to the nth case.
http://www.maths.surrey.ac.uk/hosted-sites/R.Knott/Fibonacci/fibformproof.html

The way I learned the closed form solution was to take the z-transform of the equation. Here is a link that describes how to do this.
http://mathforum.org/library/drmath/view/51540.html
The difference equation for the Fibonacci series is:
    u(n+2) = u(n+1) + u(n)   with   u(0) = 0,  u(1) = 1
So:

    u(n+2) - u(n+1) - u(n) = 0  

Taking the z-transforms:

    z^2[u(z) - u(0) - u(1)/z] - z[u(z) - u(0)] - u(z) = 0

    u(z)[z^2 - z - 1] - z^2*u(0) - z*u(1) + z*u(0) = 0

Putting u(0) = 0 and u(1) = 1 this becomes:

    u(z)[z^2 - z - 1] =  z

                 z
    u(z) =  -------------
             z^2 - z - 1

The denominator factorizes to:

    [z-1/2 -sqrt(5)/2)][z-1/2 +sqrt(5)/2]

The trick here is to express u(z)/z in partial fractions:

                                1                     1
    u(z)/z =   1/sqrt(5)[ ----------------   - ---------------- ]
                         z-1/2 - sqrt(5)/2     z-1/2 +sqrt(5)/2

and so: 

                            z                     z
    u(z) = 1/sqrt(5)[----------------  - ------------------ ]
                     z-1/2-sqrt(5)/2)      z-1/2+sqrt(5)/2

Then from table of inverse transforms:

       z
    -------  =  a^n    
     z - a

where a is constant. Then:

                              z                       z
    u(z) = 1/sqrt(5) [-------------------- - --------------------]
                       z - (1/2+sqrt(5)/2)    z - (1/2-sqrt(5)/2)

Now from the table of inverse transforms:
    u(n) = 1/sqrt(5)[(1/2+sqrt(5)/2)^n - (1/2-sqrt(5)/2)^n]

Open in new window

Here's a graph of the continuous function (i.e., let n be a continuous real positive number) that I asked wolframalpha to plot.
 u(n) = [(1/2+sqrt(5)/2)^n - (1/2-sqrt(5)/2)^n]/sqrt(5)

Open in new window

fib.pnghttp://www.wolframalpha.com/input/?i=plot+%5B(1%2F2%2Bsqrt(5)%2F2)%5En+-+(1%2F2-sqrt(5)%2F2)%5En%5D%2Fsqrt(5),+where+n+%3D+1+to+8

When n is a positive integer, then u(n) is also a positive integer. Pretty amazing!
In practice, when computing, you probably won't get a positive integer, but a real number that is very close. To get the positive integer, you need to use a round function.
http://www.tutorialspoint.com/java/number_round.htm
As n gets larger, the speed at which you compute fibonacci(n) really rocks over the other approaches, whether recursive or just looping. However, if you want a table of the first N values of fibonacci(n), then just adding the previous two integers and storing it into an array of length N will be faster than using fibonacci(n) for all values of n = 1..N, where n is an integer.

Author

Commented:
public int fibonacci(int n) {
  
   if ( n == 0 || n == 1) // base cases
      return n;
    else
      // recursion step
      return fibonacci(n - 1) + fibonacci(n - 2);
  

}

Open in new window


got rid of extra brackets. when i have to put and when i can get rid of brackets?
mccarlIT Business Systems Analyst / Software Developer
Top Expert 2015

Commented:
They make absolutely NO difference (in this case) to the execution of the code, if it is easy for you to read with them there, leave them there. That's all that it comes down to, how easy it is to read for YOU.
Yes +1 for what mccarl said - I actually think it's a little better with the extra brackets.  You should always prefer clarity and to me the extra with the extra braces is clearer.

Doug

Do more with

Expert Office
Submit tech questions to Ask the Experts™ at any time to receive solutions, advice, and new ideas from leading industry professionals.

Start 7-Day Free Trial