Recurrence sequence in Java / Python / Mathematica - java

How can you write the following statement in the given languages?
a(0) = 1
a_(n+1) = 1 - 1 / ( a_n + 3)
I need to find the smallest value of n when a_n -> 0.732050....
My attempt in Mathematica
a[(x+1)_] = 1 - 1/(a[x_] + 3)
The problem is apparently in this a[(x+1)_].
However, I do not know how to do it iteratively in Mathematica.

Mathematica
a[0] = 1;
a[n_] := a[n] = 1 - 1/(a[n-1] + 3)
(Note the memoization trick.)
Also, a[n] converges (very quickly) to sqrt(3)-1:
Solve[x == 1 - 1/(x+3), x]

Python, simplest:
def a(n):
if n == 0: return 1
return 1 - 1 / float(a(n-1) + 3)
# limit is sqrt(3) - 1
limit = 3.0 ** 0.5 - 1.0
# get 9 digits' precision
i = 0
while abs(a(i) - limit) > 1.0e-9:
i += 1
print i
This emits 8, suggesting that optimizations such as recursion elimination or memoizing are likely not warranted.
Of course normally we'd want to get the limit numerically rather than analytically, so the normal way to loop would be rather different -- and best encapsulated in a higher-order function...:
# get a function's limit numerically
def limit(f, eps=1.0e-11):
previous_value = f(0)
next_value = f(1)
i = 2
while abs(next_value - previous_value) > eps:
previous_value = next_value
next_value = f(i)
i += 1
return next_value
Nontrivial looping logic is usually best encapsulated in a generator:
def next_prev(f):
previous_value = f(0)
i = 1
while True:
next_value = f(i)
yield next_value, previous_value
i += 1
previous_value = next_value
with the help of this generator, the limit HOF becomes much simpler:
def limit(f, eps=1.0e-11):
for next_value, previous_value in next_prev(f):
if abs(next_value - previous_value) < eps:
return next_value
Note how useful the separation is: next_prev embodies the concept of "get the next and previous value of the function", limit just deals with "when should the loop terminate".
Last but not least, itertools often offers a good alternative to generators, letting you encapsulate finicky iteration logic in speedy ways (though it does take some getting used to...;-):
import itertools
def next_prev(f):
values = itertools.imap(f, itertools.count())
prv, nxt = itertools.tee(values)
nxt.next()
return itertools.izip(prv, nxt)

Java
double A = 1;
int n = 0;
while (true) {
System.out.println(n + " " + A);
A = 1 - 1 / (A + 3);
n++;
}
Python
A = 1.0
n = 0
while 1:
print n, A
A = 1 - 1 / (A + 3)
n += 1

Mathematica:
a[0] := 1
a[k_] := 1 - 1/(a[k - 1] + 3)
I substituted k = n + 1 because that makes the expression simpler. The result is equivalent.

Python
next = lambda x: 1.0 - (1.0 / (float(x) + 3.0))
last, z, count = -1, 0.0, 0
while last != z:
print count, z
last, z, count = z, next(z), count+1
I try to avoid writing "while True" or such if I can avoid it. Almost certainly no code that I write will loop forever. In this case, it ran sixteen times for me. Sixteen is a lot less than ℵ-null.

A one-liner in Mathematica which gives a list of exact elements of your sequence:
In[66]:= NestWhileList[1 - 1/(#1 + 3) &, 1,
RealExponent[Subtract[##]] > -8 &, 2]
Out[66]= {1, 3/4, 11/15, 41/56, 153/209, 571/780, 2131/2911, \
7953/10864, 29681/40545}
The difference between the last two elements is less than 10^-8. It thus have taken 8 iterations:
In[67]:= Length[%]
Out[67]= 9

Related

append strings with increasing frequency

You are given two strings S and T. An infinitely long string is formed in the following manner:
Take an empty string,
Append S one time,
Append T two times,
Append S three times,
Append T four times,
and so on, appending the strings alternately and increasing the number of repetitions by 1 each time.
You will also be given an integer K.
You need to tell the Kth Character of this infinitely long string.
Sample Input (S, T, K):
a
bc
4
Sample Output:
b
Sample Explanation:
The string formed will be "abcbcaaabcbcbcbcaaaaa...". So the 4th character is "b".
My attempt:
public class FindKthCharacter {
public char find(String S, String T, int K) {
// lengths of S and T
int s = S.length();
int t = T.length();
// Counters for S and T
int sCounter = 1;
int tCounter = 2;
// To store final chunks of string
StringBuilder sb = new StringBuilder();
// Loop until K is greater than zero
while (K > 0) {
if (K > sCounter * s) {
K -= sCounter * s;
sCounter += 2;
if (K > tCounter * t) {
K -= tCounter * t;
tCounter += 2;
} else {
return sb.append(T.repeat(tCounter)).charAt(K - 1);
}
} else {
return sb.append(S.repeat(sCounter)).charAt(K - 1);
}
}
return '\u0000';
}
}
But is there any better way to reduce its time complexity?
I've tried to give a guide here, rather than just give the solution.
If s and t are the lengths of the strings S and T, then you need to find the largest odd n such that
(1+3+5+...+n)s + (2+4+6+...+(n+1))t < K.
You can simplify these expressions to get a quadratic equation in n.
Let N be (1+3+..+n)s + (2+4+6+...+(n+1))t. You know that K will lie either in the next (n+2) copies of S, or the (n+3) copies of T that follow. Compare K to N+(n+2)s, and take the appropriate letter of either S or T using a modulo.
The only difficult step here is solving the large quadratic, but you can do it in O(log K) arithmetic operations easily enough by doubling n until it's too large, and then using a binary search on the remaining range. (If K is not too large so that floating point is viable, you can do it in O(1) time using the well-known quadratic formula).
Here my quick attempt, there probably is a better solution. Runtime is still O(sqrt n), but memory is O(1).
public static char find(String a, String b, int k) {
int lenA = a.length();
int lenB = b.length();
int rep = 0;
boolean isA = false;
while (k >= 0) {
++rep;
isA = !isA;
k -= (isA ? lenA : lenB) * rep;
}
int len = (isA ? lenA : lenB);
int idx = (len * rep + k) % len;
return (isA ? a : b).charAt(idx);
}
Here's a O(1) solution that took me some time to come up with (read I would have failed an interview on time). Hopefully the process is clear and you can implement it in code.
Our Goal is to return the char that maps to the kth index.
But How? Just 4 easy steps, actually.
Step 1: Find out how many iterations of our two patterns it would take to represent at least k characters.
Step 2: Using this above number of iterations i, return how many characters are present in the previous i-1 iterations.
Step 3: Get the number of characters n into iteration i that our kth character is. (k - result of step 2)
Step 4: Mod n by the length of the pattern to get index into pattern for the specific char. If i is odd, look into s, else look into t.
For step 1, we need to find a formula to give us the iteration i that character k is in. To derive this formula, it may be easier to first derive the formula needed for step 2.
Step 2's formula is basically given an iteration i, return how many characters are present in that iteration. We are solving for 'k' in this equation and are given i, while it's the opposite for step 1 where were are solving for i given k. If we can derive the equation of find k given i, then we can surely reverse it to find i given k.
Now, let's try to derive the formula for step 2 and find k given i. Here it's best to start with the most basic example to see the pattern.
s = "a", t = "b"
i=1 a
i=2 abb
i=3 abbaaa
i=4 abbaaabbbb
i=5 abbaaabbbbaaaaa
i=6 abbaaabbbbaaaaabbbbbb
Counting the total number of combined chars for each pattern during its next iteration gives us:
#iterations of pattern: 1 2 3 4 5 6 7 8 9 10
every new s iteration: 1, 4, 9, 16, 25, 36, 49, 64, 81, 100
every new t iteration: 2, 6, 12, 20, 30, 42, 56, 72, 90, 110
You might notice some nice patterns here. For example, s has a really nice formula to find out how many combined characters it has at any given iteration. It's simply (# of s iterations^2)*s.length. t also has a simple formula. It is (# of t iterations * (# of t iterations + 1))*t.length. You may have noticed that these formulas are the formulas for sum of odd and even numbers (if you did you get a kudos). This makes sense because each pattern's sum for an iteration i is the sum of all of its previous iterations.
Using s,t as length of their respective patterns, we now have the following formula to find the total number of chars at a given iteration.
#chars = s*(# of s iterations)^2 + t * (# of t iterations * (# of t iterations + 1))
Now we just need to do some math to get the number of iterations for each pattern given i.
# of s iterations given i = ceil(i/2.0)
# of t iterations given i = floor(i/2) which / operation gives us by default
Plugging these back into our formula we get:
total # of chars = s*(ceil(i/2.0)^2) + t*((i/2)*((i/2)+1))
We have just completed step 2, and we now know at any given iteration how many total chars there are. We could stop here and start picking random iterations and adjusting accordingly until we get near k, but we can do better than that. Let's use the above formula now to complete step 1 which we skipped. We just need to reorganize our equation to solve for i now.
Doing some simplyfying we get:
// 2
// i i i
// s (-) + t - ( - + 1 ) = k
// 2 2 2
// ----------------------------
// 2
// i t i
// s - + - ( - + 1 )i = k
// 4 2 2
// ----------------------------
// 2 2
// si ti ti
// ---- + ---- + ---- - k = 0
// 4 4 2
// ----------------------------
//
// 2 2
// si + ti + 2ti - 4k = 0
// ----------------------------
// 2
// (s + t)i + 2ti - 4k = 0
// ----------------------------
This looks like a polynomial. Wow! You're right! That means we can solve it using the quadratic formula.
A=(s+t), B=2t, C=-4k
quadratic formula = (-2t + sqrt(2t^2 + 16(s+t)k)) / 2(s+t)
This is our formula for step 1, and it will give us the iteration that the kth character is on. We just need to ceil it. I'm actually not smart enough to know why this works. It just does. Here is a desmos graph that graphs our two polynomials from step 2: s(Siterations)^2 and t(Titerations (Titerations + 1)).
The area under both curves is our total number of chars at an iteration (the vertical line). The formula from step 1 is also graphed, and we can see that for any s, t, k that the x intercept (which represents our xth iteration) is always: previous iteration < x <= current iteration, which is why the ceil works.
We have now completed steps 1 and 2. We have a formula to get the ith iteration that the kth character is on and a formula that gives us how many characters are in an ith iteration. Steps 3 and 4 should follow and we get our answer. This is constant time.

Need help writing a method that adds two polynomials together

I am currently working on writing a method that adds two polynomials (given by 2 text files) together. So for example:
4.0x^5 + -2.0x^3 + 2.0x + 3.0
&
8.0x^4 + 4.0x^3 + -3.0x + 9.0
would result in: 4.0x^5 + 8.0x^4 + 2.0x^3 - 1.0x + 12
Currently, my method creates a new polynomial object, but only adds the terms in which there is a degree match. So my output looks like this:
2.0x^3 - 1.0x + 12
It is missing the first two terms, because the degrees don't match up. Here is my code (important: Polynomial is initially constructed by Node poly = null; -- so the variable poly is a node that points to the front of the Linked List of the polynomial):
public Polynomial add(Polynomial p) {
Polynomial answer = new Polynomial();
for (Node firstPoly = poly; firstPoly != null; firstPoly = firstPoly.next){
for (Node secondPoly = p.poly; secondPoly != null; secondPoly = secondPoly.next){
if (firstPoly.term.degree == secondPoly.term.degree){
answer = addToRear(answer, (firstPoly.term.coeff + secondPoly.term.coeff), firstPoly.term.degree, null);
if (answer.poly.term.coeff == 0){
answer.poly = null;
}
}
}
}
return answer;
}
I'm not asking anyone to solve this for me, but does anyone have any idea what I would do next to make sure the non-matching degrees get added? I've been trying to work it out on paper, but nothing is working out for various reasons. Here is the addToRear method just in case it'll be useful to you guys.
private Polynomial addToRear(Polynomial p, float coeff, int degree, Node next){
if (p.poly == null){
p.poly = new Node(coeff, degree, null);
return p;
}
for (Node temp = p.poly; temp != null; temp = temp.next){
if (temp.next == null){
temp.next = new Node(coeff, degree, null);
return p;
}
}
return p;
}
Thank you.
It's probably easier to ensure that a polynomial data structure includes all powers, including those with zero multipliers. In other words, the polynomial 2.0x3 - 1.0x + 12 would be represented as the collection:
pwr 0 1 2 3
---- ---- ---- ----
{ 12.0, 1.0, 0.0, 2.0 }
Unless you're talking about large numbers of polynomials and massive powers for the high-end terms, the non-sparseness of this solution should be mostly irrelevant.
As you can see, I've also changed the order so that the x0 (constant) term is first in the list since, assuming your polynomials don't have negative powers, this will also ease the adding effort. This is because the matching powers will then have matching indexes in the collection.
So, to add together the two polynomials:
4.0x5 - 2.0x3 + 2.0x + 3.08.0x4 + 4.0x3 - 3.0x + 9.0
that would be akin to:
pwr 0 1 2 3 4 5
---- ---- ---- ---- ---- ----
{ 3.0, 2.0, 0.0, -2.0, 0.0, 4.0 }
+ { 9.0, -3.0, 0.0, 4.0, 8.0 }
----------------------------------------
= { 12.0, -1.0, 0.0, 2.0, 8.0, 4.0 }
which gives, as desired (ignoring zero multipliers when outputting)
4.0x5 + 8.0x4 + 2.0x3 - 1.0x + 12.0
If, for some reason, you must work on sparse linked lists, then it depends on whether the terms are sorted based on the power.
If they're not, you will generally have to use an algorithm like:
set poly3 to empty
# Process all in poly1, including those in poly2.
foreach term1 in poly1:
find term2 in poly2 with matching power
if none:
add (term1.coeff, term1.power) to poly3
else:
add (term1.coeff + term2.coeff, term1.power) to poly3
# Process all in poly2, but NOT in poly1.
foreach term2 in poly2:
find term1 in poly1 with matching power
if none:
add (term2.coeff, term2.power) to poly3
This will first add all powers in the first polynomial, including those that have an entry in the second.
It will then add those in the second that don't have a corresponding term in the first. That way, all terms are added correctly.
If the terms are sorted, you can make it a little more efficient by processing the lists in parallel (similar to a merge algorithm), at each stage getting the term from the list with the highest unprocessed power, something like:
set poly3 to empty
set term1 to poly1.head
set term2 to poly2.head
# Process both until at least one runs out.
while term1 != null and term2 != null:
if term1.power == term2.power:
add (term1.coeff + term2.coeff, term1.power) to poly3
term1 = term1.next
term2 = term2.next
elif term1.power > term2.power:
add (term1.coeff, term1.power) to poly3
term1 = term1.next
else:
add (term2.coeff, term2.power) to poly3
term2 = term2.next
# Process remaining single list, if any.
while term1 != null:
add (term1.coeff, term1.power) to poly3
term1 = term1.next
while term2 != null:
add (term2.coeff, term2.power) to poly3
term2 = term2.next
As a proof of concept, here's some Python code which does the sorted variant. The bulk of the code is turning a string into a "list" (actually a power-sparse array) and printing out the resultant polynomial. The meat of the solution is in the main line, starting at poly3 = []:
poly1 = '4.0x^5 - 2.0x^3 + 2.0x + 3.0'
poly2 = '8.0x^4 + 4.0x^3 + -3.0x + 9.0'
# Makes component extraction from array easier.
coeff = 0
power = 1
def baseline(s):
# Remove spaces, normalise to all '+'.
check = s + ' '
result = s
while result != check:
check = result
result = result.replace(' ','');
result = result.replace('-','+-')
result = result.replace('++','+')
# Create array of terms.
result = result.split('+')
# Make each term a coefficient/power pair.
for i in range(len(result)):
result[i] = result[i].split('^')
if len(result[i]) == 1 and result[i][coeff].endswith('x'):
result[i].append('1')
if len(result[i]) == 1 and not result[i][coeff].endswith('x'):
result[i].append('0')
if result[i][coeff].endswith('x'):
result[i][coeff] = result[i][coeff][:-1]
result[i][coeff] = float(result[i][coeff])
result[i][power] = int(result[i][power])
return result
def polyprint(s,p):
print()
print(s, p, end=':\n ')
if len(p) > 0:
print(p[0][coeff],end='')
if p[0][power] == 1:
print('x',end='')
elif p[0][power] > 1:
print('x^%d' % (p[0][power]),end='')
for i in range(1,len(p)):
if p[i][coeff] < 0:
print(' -',-p[i][coeff],end='')
else:
print(' +',p[i][coeff],end='')
if p[i][power] == 1:
print('x',end='')
elif p[i][power] > 1:
print('x^%d' % (p[i][power]),end='')
print()
# Turn polynomials into sparse (based on powers) array.
poly1 = baseline(poly1)
poly2 = baseline(poly2)
polyprint('poly1',poly1)
polyprint('poly2',poly2)
# Add them as per sorted algorithm.
poly3 = []
idx1 = 0
idx2 = 0
while idx1 < len(poly1) and idx2 < len(poly2):
if poly1[idx1][power] == poly2[idx2][power]:
if poly1[idx1][coeff] != poly2[idx2][coeff]:
poly3.append([poly1[idx1][coeff] + poly2[idx2][coeff], poly1[idx1][power]])
idx1 += 1
idx2 += 1
continue
if poly1[idx1][power] > poly2[idx2][power]:
poly3.append([poly1[idx1][coeff], poly1[idx1][power]])
idx1 += 1
continue
poly3.append([poly2[idx2][coeff], poly2[idx2][power]])
idx2 += 1
while idx1 < len(poly1):
poly3.append([poly1[idx1][coeff], poly1[idx1][power]])
idx1 += 1
while idx2 < len(poly2):
poly3.append([poly2[idx2][coeff], poly2[idx2][power]])
idx2 += 1
polyprint('poly3',poly3)
Take the maximum degree of either polynomial, and set all missing terms to zero. That is, x^2 = 0x^3 + 1x^2 + 0x^1 + 0x^0. (As a side note, a real implementation for numeric computing would represent polynomials in nested form, to minimize the loss of precision from adding after multiplication.)

Need help in understanding Rolling Hash computation in constant time for Rabin-Karp Implementation

I've been trying to implement Rabin-Karp algorithm in Java. I have hard time computing the rolling hash value in constant time. I've found one implementation at http://algs4.cs.princeton.edu/53substring/RabinKarp.java.html. Still I could not get how these two lines work.
txtHash = (txtHash + Q - RM*txt.charAt(i-M) % Q) % Q;
txtHash = (txtHash*R + txt.charAt(i)) % Q;
I looked at couple of articles on modular arithmetic but no article could able to penetrate my thick skull. Please give some pointers to understand this.
First you need to understand how the hash is computed.
Lets take a simple case of base 10 strings. How would you guarantee that the hash code of a string is unique? Base 10 is what we use to represent numbers, and we don't have collisions!!
"523" = 5*10^2 + 2*10^1 + 3*10^0 = 523
using the above hash function you are guaranteed to get distinct hashes for every string.
Given the hash of "523", if you want to calculate the hash of "238", i.e. by jutting out the leftmost digit 5 and bringing in a new digit 8 from the right, you would have to do the following:
1) remove the effect of the 5 from the hash:
hash = hash - 5*10^2 (523-500 = 23)
2) adjust the hash of the remaining chars by shifting by 1, hash = hash * 10
3) add the hash of the new character:
hash = hash + 8 (230 + 8 = 238, which as we expected is the base 10 hash of "238")
Now let's extend this to all ascii characters. This takes us to the base 256 world. Therefore the hash of the same string "523" now is
= 5*256^2 + 2*256^1 + 3*256^0 = 327680 + 512 + 3 = 328195.
You can imagine as the string length increases you will will exceed the range of integer/long in most programming languages relatively quickly.
How can we solve this? The way this is routinely solved is by working with modulus a large prime number. The drawback of this method is that we will now get false positives as well, which is a small price to pay if it takes the runtime of your algorithm from quadratic to linear!
The complicated equation you quoted is nothing but the steps 1-3 above done with modulus math.
The two modulus properties used above are ->
a) (a*b) % p = ((a % p) * (b % p)) % p
b) a % p = (a + p) % p
Lets go back to steps 1-3 mentioned above ->
1) (expanded using property a) hash = hash - ((5 % p)*(10^2 %p) %p)
vs. what you quoted
txtHash = (txtHash + Q - RM*txt.charAt(i-M) % Q) % Q;
Here are is how the two are related!
RM = 10^3 % p
txt.charAt(i-M) % Q = 5 % p
The additional + Q you see is just to ensure that the hash is not negative. See property b above.
2 & 3) hash = hash*10 + 8, vs txtHash = (txtHash*R + txt.charAt(i)) % Q;
Is the same but with taking mod of the final hash result!
Looking at properties a & b more closely, should help you figure it out!
This is the "rolling" aspect of the hash. It's eliminating the contribution of the oldest character (txt.charAt(i-M)), and incorporating the contribution of the newest character(txt.charAt(i)).
The hash function is defined as:
M-1
hash[i] = ( SUM { input[i-j] * R^j } ) % Q
j=0
(where I'm using ^ to denote "to the power of".)
But this can be written as an efficient recursive implementation as:
hash[i] = (txtHash*R - input[i-M]*(R^M) + input[i]) % Q
Your reference code is doing this, but it's using various techniques to ensure that the result is always computed correctly (and efficiently).
So, for instance, the + Q in the first expression has no mathematical effect, but it ensures that the result of the sum is always positive (if it goes negative, % Q doesn't have the desired effect). It's also breaking the calculation into stages, presumably to prevent numerical overflow.

Bitwise Multiply and Add in Java

I have the methods that do both the multiplication and addition, but I'm just not able to get my head around them. Both of them are from external websites and not my own:
public static void bitwiseMultiply(int n1, int n2) {
int a = n1, b = n2, result=0;
while (b != 0) // Iterate the loop till b==0
{
if ((b & 01) != 0) // Logical ANDing of the value of b with 01
{
result = result + a; // Update the result with the new value of a.
}
a <<= 1; // Left shifting the value contained in 'a' by 1.
b >>= 1; // Right shifting the value contained in 'b' by 1.
}
System.out.println(result);
}
public static void bitwiseAdd(int n1, int n2) {
int x = n1, y = n2;
int xor, and, temp;
and = x & y;
xor = x ^ y;
while (and != 0) {
and <<= 1;
temp = xor ^ and;
and &= xor;
xor = temp;
}
System.out.println(xor);
}
I tried doing a step-by-step debug, but it really didn't make much sense to me, though it works.
What I'm possibly looking for is to try and understand how this works (the mathematical basis perhaps?).
Edit: This is not homework, I'm just trying to learn bitwise operations in Java.
Let's begin by looking the multiplication code. The idea is actually pretty clever. Suppose that you have n1 and n2 written in binary. Then you can think of n1 as a sum of powers of two: n2 = c30 230 + c29 229 + ... + c1 21 + c0 20, where each ci is either 0 or 1. Then you can think of the product n1 n2 as
n1 n2 =
n1 (c30 230 + c29 229 + ... + c1 21 + c0 20) =
n1 c30 230 + n1 c29 229 + ... + n1 c1 21 + n1 c0 20
This is a bit dense, but the idea is that the product of the two numbers is given by the first number multiplied by the powers of two making up the second number, times the value of the binary digits of the second number.
The question now is whether we can compute the terms of this sum without doing any actual multiplications. In order to do so, we're going to need to be able to read the binary digits of n2. Fortunately, we can do this using shifts. In particular, suppose we start off with n2 and then just look at the last bit. That's c0. If we then shift the value down one position, then the last bit is c0, etc. More generally, after shifting the value of n2 down by i positions, the lowest bit will be ci. To read the very last bit, we can just bitwise AND the value with the number 1. This has a binary representation that's zero everywhere except the last digit. Since 0 AND n = 0 for any n, this clears all the topmost bits. Moreover, since 0 AND 1 = 0 and 1 AND 1 = 1, this operation preserves the last bit of the number.
Okay - we now know that we can read the values of ci; so what? Well, the good news is that we also can compute the values of the series n1 2i in a similar fashion. In particular, consider the sequence of values n1 << 0, n1 << 1, etc. Any time you do a left bit-shift, it's equivalent to multiplying by a power of two. This means that we now have all the components we need to compute the above sum. Here's your original source code, commented with what's going on:
public static void bitwiseMultiply(int n1, int n2) {
/* This value will hold n1 * 2^i for varying values of i. It will
* start off holding n1 * 2^0 = n1, and after each iteration will
* be updated to hold the next term in the sequence.
*/
int a = n1;
/* This value will be used to read the individual bits out of n2.
* We'll use the shifting trick to read the bits and will maintain
* the invariant that after i iterations, b is equal to n2 >> i.
*/
int b = n2;
/* This value will hold the sum of the terms so far. */
int result = 0;
/* Continuously loop over more and more bits of n2 until we've
* consumed the last of them. Since after i iterations of the
* loop b = n2 >> i, this only reaches zero once we've used up
* all the bits of the original value of n2.
*/
while (b != 0)
{
/* Using the bitwise AND trick, determine whether the ith
* bit of b is a zero or one. If it's a zero, then the
* current term in our sum is zero and we don't do anything.
* Otherwise, then we should add n1 * 2^i.
*/
if ((b & 1) != 0)
{
/* Recall that a = n1 * 2^i at this point, so we're adding
* in the next term in the sum.
*/
result = result + a;
}
/* To maintain that a = n1 * 2^i after i iterations, scale it
* by a factor of two by left shifting one position.
*/
a <<= 1;
/* To maintain that b = n2 >> i after i iterations, shift it
* one spot over.
*/
b >>>= 1;
}
System.out.println(result);
}
Hope this helps!
It looks like your problem is not java, but just calculating with binary numbers. Start of simple:
(all numbers binary:)
0 + 0 = 0 # 0 xor 0 = 0
0 + 1 = 1 # 0 xor 1 = 1
1 + 0 = 1 # 1 xor 0 = 1
1 + 1 = 10 # 1 xor 1 = 0 ( read 1 + 1 = 10 as 1 + 1 = 0 and 1 carry)
Ok... You see that you can add two one digit numbers using the xor operation. With an and you can now find out whether you have a "carry" bit, which is very similar to adding numbers with pen&paper. (Up to this point you have something called a Half-Adder). When you add the next two bits, then you also need to add the carry bit to those two digits. Taking this into account you can get a Full-Adder. You can read about the concepts of Half-Adders and Full-Adders on Wikipedia:
http://en.wikipedia.org/wiki/Adder_(electronics)
And many more places on the web.
I hope that gives you a start.
With multiplication it is very similar by the way. Just remember how you did multiplying with pen&paper in elementary school. Thats what is happening here. Just that it's happening with binary numbers and not with decimal numbers.
EXPLANATION OF THE bitwiseAdd METHOD:
I know this question was asked a while back but since no complete answer has been given regarding how the bitwiseAdd method works here is one.
The key to understanding the logic encapsulated in bitwiseAdd is found in the relationship between addition operations and xor and and bitwise operations. That relationship is defined by the following equation (see appendix 1 for a numeric example of this equation):
x + y = 2 * (x&y)+(x^y) (1.1)
Or more simply:
x + y = 2 * and + xor (1.2)
with
and = x & y
xor = x ^ y
You might have noticed something familiar in this equation: the and and xor variables are the same as those defined at the beginning of bitwiseAdd. There is also a multiplication by two, which in bitwiseAdd is done at the beginning of the while loop. But I will come back to that later.
Let me also make a quick side note about the '&' bitwise operator before we proceed further. This operator basically "captures" the intersection of the bit sequences against which it is applied. For example, 9 & 13 = 1001 & 1101 = 1001 (=9). You can see from this result that only those bits common to both bit sequences are copied to the result. It derives from this that when two bit sequences have no common bit, the result of applying '&' on them yields 0. This has an important consequence on the addition-bitwise relationship which shall become clear soon
Now the problem we have is that equation 1.2 uses the '+' operator whereas bitwiseAdd doesn't (it only uses '^', '&' and '<<'). So how do we make the '+' in equation 1.2 somehow disappear? Answer: by 'forcing' the and expression to return 0. And the way we do that is by using recursion.
To demonstrate this I am going to recurse equation 1.2 one time (this step might be a bit challenging at first but if needed there's a detailed step by step result in appendix 2):
x + y = 2*(2*and & xor) + (2*and ^ xor) (1.3)
Or more simply:
x + y = 2 * and[1] + xor[1] (1.4)
with
and[1] = 2*and & xor,
xor[1] = 2*and ^ xor,
[1] meaning 'recursed one time'
There's a couple of interesting things to note here. First you noticed how the concept of recursion sounds close to that of a loop, like the one found in bitwiseAdd in fact. This connection becomes even more obvious when you consider what and[1] and xor[1] are: they are the same expressions as the and and xor expressions defined INSIDE the while loop in bitwiseAdd. We also note that a pattern emerges: equation 1.4 looks exactly like equation 1.2!
As a result of this, doing further recursions is a breeze, if one keeps the recursive notation. Here we recurse equation 1.2 two more times:
x + y = 2 * and[2] + xor[2]
x + y = 2 * and[3] + xor[3]
This should now highlight the role of the 'temp' variable found in bitwiseAdd: temp allows to pass from one recursion level to the next.
We also notice the multiplication by two in all those equations. As mentioned earlier this multiplication is done at the begin of the while loop in bitwiseAdd using the and <<= 1 statement. This multiplication has a consequence on the next recursion stage since the bits in and[i] are different from those in the and[i] of the previous stage (and if you recall the little side note I made earlier about the '&' operator you probably see where this is going now).
The general form of equation 1.4 now becomes:
x + y = 2 * and[x] + xor[x] (1.5)
with x the nth recursion
FINALY:
So when does this recursion business end exactly?
Answer: it ends when the intersection between the two bit sequences in the and[x] expression of equation 1.5 returns 0. The equivalent of this in bitwiseAdd happens when the while loop condition becomes false. At this point equation 1.5 becomes:
x + y = xor[x] (1.6)
And that explains why in bitwiseAdd we only return xor at the end!
And we are done! A pretty clever piece of code this bitwiseAdd I must say :)
I hope this helped
APPENDIX:
1) A numeric example of equation 1.1
equation 1.1 says:
x + y = 2(x&y)+(x^y) (1.1)
To verify this statement one can take a simple example, say adding 9 and 13 together. The steps are shown below (the bitwise representations are in parenthesis):
We have
x = 9 (1001)
y = 13 (1101)
And
x + y = 9 + 13 = 22
x & y = 9 & 13 = 9 (1001 & 1101 = 1001)
x ^ y = 9^13 = 4 (1001 ^ 1101 = 0100)
pluging that back into equation 1.1 we find:
9 + 13 = 2 * 9 + 4 = 22 et voila!
2) Demonstrating the first recursion step
The first recursion equation in the presentation (equation 1.3) says that
if
x + y = 2 * and + xor (equation 1.2)
then
x + y = 2*(2*and & xor) + (2*and ^ xor) (equation 1.3)
To get to this result, we simply took the 2* and + xor part of equation 1.2 above and applied the addition/bitwise operands relationship given by equation 1.1 to it. This is demonstrated as follow:
if
x + y = 2(x&y) + (x^y) (equation 1.1)
then
[2(x&y)] + (x^y) = 2 ([2(x&y)] & (x^y)) + ([2(x&y)] ^ (x^y))
(left side of equation 1.1) (after applying the addition/bitwise operands relationship)
Simplifying this with the definitions of the and and xor variables of equation 1.2 gives equation 1.3's result:
[2(x&y)] + (x^y) = 2*(2*and & xor) + (2*and ^ xor)
with
and = x&y
xor = x^y
And using that same simplification gives equation 1.4's result:
2*(2*and & xor) + (2*and ^ xor) = 2*and[1] + xor[1]
with
and[1] = 2*and & xor
xor[1] = 2*and ^ xor
[1] meaning 'recursed one time'
Here is another approach for Multiplication
/**
* Multiplication of binary numbers without using '*' operator
* uses bitwise Shifting/Anding
*
* #param n1
* #param n2
*/
public static void multiply(int n1, int n2) {
int temp, i = 0, result = 0;
while (n2 != 0) {
if ((n2 & 1) == 1) {
temp = n1;
// result += (temp>>=(1/i)); // To do it only using Right shift
result += (temp<<=i); // Left shift (temp * 2^i)
}
n2 >>= 1; // Right shift n2 by 1.
i++;
}
System.out.println(result);
}

Best way to make Java's modulus behave like it should with negative numbers?

In java when you do
a % b
If a is negative, it will return a negative result, instead of wrapping around to b like it should. What's the best way to fix this? Only way I can think is
a < 0 ? b + a : a % b
It behaves as it should a % b = a - a / b * b; i.e. it's the remainder.
You can do (a % b + b) % b
This expression works as the result of (a % b) is necessarily lower than b, no matter if a is positive or negative. Adding b takes care of the negative values of a, since (a % b) is a negative value between -b and 0, (a % b + b) is necessarily lower than b and positive. The last modulo is there in case a was positive to begin with, since if a is positive (a % b + b) would become larger than b. Therefore, (a % b + b) % b turns it into smaller than b again (and doesn't affect negative a values).
As of Java 8, you can use Math.floorMod(int x, int y) and Math.floorMod(long x, long y). Both of these methods return the same results as Peter's answer.
Math.floorMod( 2, 3) = 2
Math.floorMod(-2, 3) = 1
Math.floorMod( 2, -3) = -1
Math.floorMod(-2, -3) = -2
For those not using (or not able to use) Java 8 yet, Guava came to the rescue with IntMath.mod(), available since Guava 11.0.
IntMath.mod( 2, 3) = 2
IntMath.mod(-2, 3) = 1
One caveat: unlike Java 8's Math.floorMod(), the divisor (the second parameter) cannot be negative.
In number theory, the result is always positive. I would guess that this is not always the case in computer languages because not all programmers are mathematicians. My two cents, I would consider it a design defect of the language, but you can't change it now.
=MOD(-4,180) = 176
=MOD(176, 180) = 176
because 180 * (-1) + 176 = -4 the same as 180 * 0 + 176 = 176
Using the clock example here, http://mathworld.wolfram.com/Congruence.html
you would not say duration_of_time mod cycle_length is -45 minutes, you would say 15 minutes, even though both answers satisfy the base equation.
Java 8 has Math.floorMod, but it is very slow (its implementation has multiple divisions, multiplications, and a conditional). Its possible that the JVM has an intrinsic optimized stub for it, however, which would speed it up significantly.
The fastest way to do this without floorMod is like some other answers here, but with no conditional branches and only one slow % op.
Assuming n is positive, and x may be anything:
int remainder = (x % n); // may be negative if x is negative
//if remainder is negative, adds n, otherwise adds 0
return ((remainder >> 31) & n) + remainder;
The results when n = 3:
x | result
----------
-4| 2
-3| 0
-2| 1
-1| 2
0| 0
1| 1
2| 2
3| 0
4| 1
If you only need a uniform distribution between 0 and n-1 and not the exact mod operator, and your x's do not cluster near 0, the following will be even faster, as there is more instruction level parallelism and the slow % computation will occur in parallel with the other parts as they do not depend on its result.
return ((x >> 31) & (n - 1)) + (x % n)
The results for the above with n = 3:
x | result
----------
-5| 0
-4| 1
-3| 2
-2| 0
-1| 1
0| 0
1| 1
2| 2
3| 0
4| 1
5| 2
If the input is random in the full range of an int, the distribution of both two solutions will be the same. If the input clusters near zero, there will be too few results at n - 1 in the latter solution.
Here is an alternative:
a < 0 ? b-1 - (-a-1) % b : a % b
This might or might not be faster than that other formula [(a % b + b) % b]. Unlike the other formula, it contains a branch, but uses one less modulo operation. Probably a win if the computer can predict a < 0 correctly.
(Edit: Fixed the formula.)

Categories