I've tried to find a good way to speed up the code for a problem I've been working on. The basic idea of the code is very simple. There are five inputs:
Four 1xm (for some m < n, they can be different sizes) matrices (A, B, C, D) that are pairwise-disjoint subsets of {1,2,...,n} and one nxn symmetric binary matrix (M). The basic idea for the code is to check an inequality for for every combination of elements and if the inequality holds, return the values that cause it to hold, i.e.:
for a = A
for b = B
for c = C
for d = D
if M(a,c) + M(b,d) < M(a,d) + M(b,c)
result = [a b c d];
return
end
end
end
end
end
I know there has to be a better way to do this. First, since it's symmetric, I can cut down half of the items checked since M(a,b) = M(b,a). I've been researching vectorization, found several functions I'd never heard of with MATLAB (since I'm relatively new), but I can't find anything that will particularly help me with this specific problem. I've thought of other ways to approach the problem, but nothing has been perfected, and I just don't know what to do at this point.
For example, I could possibly split this into two cases:
1) The right hand side is 1: then I have to check that both terms on the left side are 0.
2) The right hand side is 2: then I have to check that at least one term on the left hand side is 0.
But, again, I won't be able to avoid nesting.
I appreciate all the help you can offer. Thank you!
You're asking two questions here: (1) is there a more efficient algorithm to perform this search, and (2) how can I vectorize this in MATLAB. The first one is very interesting to think about, but may be a little beyond the scope of this forum. The second one is easier to answer.
As pointed out in the comments below your question, you can vectorize the for loop by enumerating all of the possibilities and checking them all together, and the answers from this question can help:
[a,b,c,d] = ndgrid(A,B,C,D); % Enumerate all combos
a=a(:); b=b(:); c=c(:); d=d(:); % Reshape from 4-D matrices to vectors
ac = sub2ind(size(M),a,c); % Convert subscript pairs to linear indices
bd = sub2ind(size(M),b,d);
ad = sub2ind(size(M),a,d);
bc = sub2ind(size(M),b,c);
mask = (M(ac) + M(bd) < M(ad) + M(bc)); % Test the inequality
results = [a(mask), b(mask), c(mask), d(mask)]; % Select the ones that pass
Again, this isn't an algorithmic change: it still has the same complexity as your nested for loop. The vectorization may cause it to run faster, but it also lacks early termination, so in certain cases it may be slower.
Since M is binary, we can think about this as a graph problem. i,j in {1..n} correspond to nodes, and M(i,j) indicates whether there is an undirected edge connecting them.
Since A,B,C,D are disjoint, that simplifies the problem a bit. We can approach the problem in stages:
Find all (c,d) for which there exists a such that M(a,c) < M(a,d). Let's call this set CD_lt_a, (the subset of C*D such that the "less than" inequality holds for some a).
Find all (c,d) for which there exists a such that M(a,c) <= M(a,d), and call this set CD_le_a.
Repeat for b, forming CD_lt_b for M(b,d) < M(b,c) and CD_le_b for M(b,d)<=M(b,c).
One way to satisfy the overall inequality is for M(a,c) < M(a,d) and M(b,d) <= M(b,c), so we can look at the intersection of CD_lt_a and CD_le_b.
The other way is if M(a,c) <= M(a,d) and M(b,d) < M(b,c), so look at the intersection of CD_le_a and CD_lt_b.
With (c,d) known, we can go back and find the (a,b).
And so my implementation is:
% 0. Some preliminaries
% Get the size of each set
mA = numel(A); mB = numel(B); mC = numel(C); mD = numel(D);
% 1. Find all (c,d) for which there exists a such that M(a,c) < M(a,d)
CA_linked = M(C,A);
AD_linked = M(A,D);
CA_not_linked = ~CA_linked;
% Multiplying these matrices tells us, for each (c,d), how many nodes
% in A satisfy this M(a,c)<M(a,d) inequality
% Ugh, we need to cast to double to use the matrix multiplication
CD_lt_a = (CA_not_linked * double(AD_linked)) > 0;
% 2. For M(a,c) <= M(a,d), check that the converse is false for some a
AD_not_linked = ~AD_linked;
CD_le_a = (CA_linked * double(AD_not_linked)) < mA;
% 3. Repeat for b
CB_linked = M(C,B);
BD_linked = M(B,D);
CD_lt_b = (CB_linked * double(~BD_linked)) > 0;
CD_le_b = (~CB_linked * double(BD_linked)) < mB;
% 4. Find the intersection of CD_lt_a and CD_le_b - this is one way
% to satisfy the inequality M(a,c)+M(b,d) < M(a,d)+M(b,c)
CD_satisfy_ineq_1 = CD_lt_a & CD_le_b;
% 5. The other way to satisfy the inequality is CD_le_a & CD_lt_b
CD_satisfy_ineq_2 = CD_le_a & CD_lt_b;
inequality_feasible = any(CD_satisfy_ineq_1(:) | CD_satisfy_ineq_2(:));
Note that you can stop here if feasibility is your only concern. The complexity is A*C*D + B*C*D, which is better than the worst-case A*B*C*D complexity of the for loop. However, early termination means your nested for loops may still be faster in certain cases.
The next block of code enumerates all the a,b,c,d that satisfy the inequality. It's not very well optimized (it appends to a matrix from within a loop), so it can be pretty slow if there are many results.
% 6. With (c,d) known, find a and b
% We can define these functions to help us search
find_a_lt = #(c,d) find(CA_not_linked(c,:)' & AD_linked(:,d));
find_a_le = #(c,d) find(CA_not_linked(c,:)' | AD_linked(:,d));
find_b_lt = #(c,d) find(CB_linked(c,:)' & ~BD_linked(:,d));
find_b_le = #(c,d) find(CB_linked(c,:)' | ~BD_linked(:,d));
% I'm gonna assume there aren't too many results, so I will be appending
% to an array inside of a for loop. Bad for performance, but maybe a bit
% more readable for a StackOverflow answer.
results = zeros(0,4);
% Find those that satisfy it the first way
[c_list,d_list] = find(CD_satisfy_ineq_1);
for ii = 1:numel(c_list)
c = c_list(ii); d = d_list(ii);
a = find_a_lt(c,d);
b = find_b_le(c,d);
% a,b might be vectors, in which case all combos are valid
% Many ways to find all combos, gonna use ndgrid()
[a,b] = ndgrid(a,b);
% Append these to the growing list of results
abcd = [a(:), b(:), repmat([c d],[numel(a),1])];
results = [results; abcd];
end
% Repeat for the second way
[c_list,d_list] = find(CD_satisfy_ineq_2);
for ii = 1:numel(c_list)
c = c_list(ii); d = d_list(ii);
a = find_a_le(c,d);
b = find_b_lt(c,d);
% a,b might be vectors, in which case all combos are valid
% Many ways to find all combos, gonna use ndgrid()
[a,b] = ndgrid(a,b);
% Append these to the growing list of results
abcd = [a(:), b(:), repmat([c d],[numel(a),1])];
results = [results; abcd];
end
% Remove duplicates
results = unique(results, 'rows');
% And actually these a,b,c,d will be indices into A,B,C,D because they
% were obtained from calling find() on submatrices of M.
if ~isempty(results)
results(:,1) = A(results(:,1));
results(:,2) = B(results(:,2));
results(:,3) = C(results(:,3));
results(:,4) = D(results(:,4));
end
I tested this on the following test case:
m = 1000;
A = (1:m); B = A(end)+(1:m); C = B(end)+(1:m); D = C(end)+(1:m);
M = rand(D(end),D(end)) < 1e-6; M = M | M';
I like to think that first part (see if the inequality is feasible for any a,b,c,d) worked pretty well. The other vectorized answers (that use ndgrid or combvec to enumerate all combinations of a,b,c,d) would require 8 terabytes of memory for a problem of this size!
But I would not recommend running the second part (enumerating all of the results) when there are more than a few hundred c,d that satisfy the inequality, because it will be pretty damn slow.
P.S. I know I answered already, but that answer was about vectorizing such loops in general, and is less specific to your particular problem.
P.P.S. This kinda reminds me of the stable marriage problem. Perhaps some of those references would contain algorithms relevant to your problem as well. I suspect that a true graph-based algorithm could probably achieve the worst-case complexity as this while additionally offering early termination. But I think it would be difficult to implement a graph-based algorithm efficiently in MATLAB.
P.P.P.S. If you only want one of the feasible solutions, you can simplify step 6 to only return a single value, e.g.
find_a_lt = #(c,d) find(CA_not_linked(c,:)' & AD_linked(:,d), 1, 'first');
find_a_le = #(c,d) find(CA_not_linked(c,:)' | AD_linked(:,d), 1, 'first');
find_b_lt = #(c,d) find(CB_linked(c,:)' & ~BD_linked(:,d), 1, 'first');
find_b_le = #(c,d) find(CB_linked(c,:)' | ~BD_linked(:,d), 1, 'first');
if any(CD_satisfy_ineq_1)
[c,d] = find(CD_satisfy_ineq_1, 1, 'first');
a = find_a_lt(c,d);
b = find_a_le(c,d);
result = [A(a), B(b), C(c), D(d)];
elseif any(CD_satisfy_ineq_2)
[c,d] = find(CD_satisfy_ineq_2, 1, 'first');
a = find_a_le(c,d);
b = find_a_lt(c,d);
result = [A(a), B(b), C(c), D(d)];
else
result = zeros(0,4);
end
If you have access to the Neural Network Toolbox, combvec could be helpful here.
running allCombs = combvec(A,B,C,D) will give you a (4 by m1*m2*m3*m4) matrix that looks like:
[...
a1, a1, a1, a1, a1 ... a1... a2... am1;
b1, b1, b1, b1, b1 ... b2... b1... bm2;
c1, c1, c1, c1, c2 ... c1... c1... cm3;
d1, d2, d3, d4, d1 ... d1... d1... dm4]
You can then use sub2ind and Matrix Indexing to setup the two values you need for your inequality:
indices = [sub2ind(size(M),allCombs(1,:),allCombs(3,:));
sub2ind(size(M),allCombs(2,:),allCombs(4,:));
sub2ind(size(M),allCombs(1,:),allCombs(4,:));
sub2ind(size(M),allCombs(2,:),allCombs(3,:))];
testValues = M(indices);
testValues(5,:) = (testValues(1,:) + testValues(2,:) < testValues(3,:) + testValues(4,:))
Your final a,b,c,d indices could be retrieved by saying
allCombs(:,find(testValues(5,:)))
Which would print a matrix with all columns which the inequality was true.
This article might be of some use.
Basically, I would like help designing an algorithm that takes a given number, and returns a random number that is unrelated to the first number. The stipulations being that a) the given output number will always be the same for a similar input number, and b) within a certain range (ex. 1-100), all output numbers are distinct. ie., no two different input numbers under 100 will give the same output number.
I know it's easy to do by creating an ordered list of numbers, shuffling them randomly, and then returning the input's index. But I want to know if it can be done without any caching at all. Perhaps with some kind of hashing algorithm? Mostly the reason for this is that if the range of possible outputs were much larger, say 10000000000, then it would be ludicrous to generate an entire range of numbers and then shuffle them randomly, if you were only going to get a few results out of it.
Doesn't matter what language it's done in, I just want to know if it's possible. I've been thinking about this problem for a long time and I can't think of a solution besides the one I've already come up with.
Edit: I just had another idea; it would be interesting to have another algorithm that returned the reverse of the first one. Whether or not that's possible would be an interesting challenge to explore.
This sounds like a non-repeating random number generator. There are several possible approaches to this.
As described in this article, we can generate them by selecting a prime number p and satisfies p % 4 = 3 that is large enough (greater than the maximum value in the output range) and generate them this way:
int randomNumberUnique(int range_len , int p , int x)
if(x * 2 < p)
return (x * x) % p
else
return p - (x * x) % p
This algorithm will cover all values in [0 , p) for an input in range [0 , p).
Here's an example in C#:
private void DoIt()
{
const long m = 101;
const long x = 387420489; // must be coprime to m
var multInv = MultiplicativeInverse(x, m);
var nums = new HashSet<long>();
for (long i = 0; i < 100; ++i)
{
var encoded = i*x%m;
var decoded = encoded*multInv%m;
Console.WriteLine("{0} => {1} => {2}", i, encoded, decoded);
if (!nums.Add(encoded))
{
Console.WriteLine("Duplicate");
}
}
}
private long MultiplicativeInverse(long x, long modulus)
{
return ExtendedEuclideanDivision(x, modulus).Item1%modulus;
}
private static Tuple<long, long> ExtendedEuclideanDivision(long a, long b)
{
if (a < 0)
{
var result = ExtendedEuclideanDivision(-a, b);
return Tuple.Create(-result.Item1, result.Item2);
}
if (b < 0)
{
var result = ExtendedEuclideanDivision(a, -b);
return Tuple.Create(result.Item1, -result.Item2);
}
if (b == 0)
{
return Tuple.Create(1L, 0L);
}
var q = a/b;
var r = a%b;
var rslt = ExtendedEuclideanDivision(b, r);
var s = rslt.Item1;
var t = rslt.Item2;
return Tuple.Create(t, s - q*t);
}
That generates numbers in the range 0-100, from input in the range 0-100. Each input results in a unique output.
It also shows how to reverse the process, using the multiplicative inverse.
You can extend the range by increasing the value of m. x must be coprime with m.
Code cribbed from Eric Lippert's article, A practical use of multiplicative inverses, and a few of the previous articles in that series.
You can not have completely unrelated (particularly if you want the reverse as well).
There is a concept of modulo inverse of a number, but this would work only if the range number is a prime, eg. 100 will not work, you would need 101 (a prime). This can provide you a pseudo random number if you want.
Here is the concept of modulo inverse:
If there are two numbers a and b, such that
(a * b) % p = 1
where p is any number, then
a and b are modular inverses of each other.
For this to be true, if we have to find the modular inverse of a wrt a number p, then a and p must be co-prime, ie. gcd(a,p) = 1
So, for all numbers in a range to have modular inverses, the range bound must be a prime number.
A few outputs for range bound 101 will be:
1 == 1
2 == 51
3 == 34
4 == 76
etc.
EDIT:
Hey...actually you know, you can use the combined approach of modulo inverse and the method as defined by #Paul. Since every pair will be unique and all numbers will be covered, your random number can be:
random(k) = randomUniqueNumber(ModuloInverse(k), p) //this is Paul's function
I have a problem that I am trying to solve but am having very little luck getting started. Here is the problem:
"If two sequences a1, a2,..., am and b1, b2,..., bn are interleaved, we say that the resulting sequence c1, c2,..., cm+n is a shuffle of the first two. For example,
DCCDBDADCACDBACB
is a shuffle of DCBDAACBB and CDDCDAC since it can be obtained by interleaving those two sequences in this way:
DC BDA AC B B
CD DC D AC"
I am to find a dynamic programming solution that figures out whether the two given subsequences are able to be merged to form the larger sequence. I have already done a problem finding the longest common subsquence and such but I can't seem to figure this one out. I know how I would go about doing it if I were to not consider dynamic programming, but I can't seem to think of how to do it with it.
Any help would be greatly appreciated.
Thank you!
You can construct a DP algorithm for solving this but first creating a recursive solution for example:
a = 'DCBDAACBB'
b = 'CDDCDAC'
c = 'DCCDBDADCACDBACB'
an = len(a)
bn = len(b)
cn = len(c)
# recursive solution O(2^n)
def isPossible(ai, bi, ci):
if ai == an and bi == bn and ci == cn:
return True
K = False
if ci < cn and ai < an and c[ci] == a[ai]:
K = K or isPossible(ai+1, bi, ci+1)
if ci < cn and bi < bn and c[ci] == b[bi]:
K = K or isPossible(ai, bi+1, ci+1)
return K
print isPossible(0, 0, 0)
Here the state can be encoded as three numbers ai, bi, ci which indicate the index of the beginning of the suffix of the original strings and isPossible(ai, bi, ci) computes whether the suffixes ai and bi can be merged into the suffix ci, we seek isPossible(0, 0, 0).
From here we can create the following DP recurrence, first initialize:
isPossible[ai][bi][ci] = False
isPossible[ai][bi][ci] = True where ai == an and bi == bn and ci == cn
Then compute:
isPossible[ai][bi][ci] = isPossible[ai+1][bi][ci+1] if A[ai] == C[ai]
isPossible[ai][bi][ci] = isPossible[ai][bi+1][ci+1] if B[ai] == C[ai]
The solution is then isPossible[0][0][0]. This runs in n^3, where as the recursive solution was 2^n
I'm trying to figure out a way to create random numbers that "feel" random over short sequences. This is for a quiz game, where there are four possible choices, and the software needs to pick one of the four spots in which to put the correct answer before filling in the other three with distractors.
Obviously, arc4random % 4 will create more than sufficiently random results over a long sequence, but in a short sequence its entirely possible (and a frequent occurrence!) to have five or six of the same number come back in a row. This is what I'm aiming to avoid.
I also don't want to simply say "never pick the same square twice," because that results in only three possible answers for every question but the first. Currently I'm doing something like this:
bool acceptable = NO;
do {
currentAnswer = arc4random() % 4;
if (currentAnswer == lastAnswer) {
if (arc4random() % 4 == 0) {
acceptable = YES;
}
} else {
acceptable = YES;
}
} while (!acceptable);
Is there a better solution to this that I'm overlooking?
If your question was how to compute currentAnswer using your example's probabilities non-iteratively, Guffa has your answer.
If the question is how to avoid random-clustering without violating equiprobability and you know the upper bound of the length of the list, then consider the following algorithm which is kind of like un-sorting:
from random import randrange
# randrange(a, b) yields a <= N < b
def decluster():
for i in range(seq_len):
j = (i + 1) % seq_len
if seq[i] == seq[j]:
i_swap = randrange(i, seq_len) # is best lower bound 0, i, j?
if seq[j] != seq[i_swap]:
print 'swap', j, i_swap, (seq[j], seq[i_swap])
seq[j], seq[i_swap] = seq[i_swap], seq[j]
seq_len = 20
seq = [randrange(1, 5) for _ in range(seq_len)]; print seq
decluster(); print seq
decluster(); print seq
where any relation to actual working Python code is purely coincidental. I'm pretty sure the prior-probabilities are maintained, and it does seem break clusters (and occasionally adds some). But I'm pretty sleepy so this is for amusement purposes only.
You populate an array of outcomes, then shuffle it, then assign them in that order.
So for just 8 questions:
answer_slots = [0,0,1,1,2,2,3,3]
shuffle(answer_slots)
print answer_slots
[1,3,2,1,0,2,3,0]
To reduce the probability for a repeated number by 25%, you can pick a random number between 0 and 3.75, and then rotate it so that the 0.75 ends up at the previous answer.
To avoid using floating point values, you can multiply the factors by four:
Pseudo code (where / is an integer division):
currentAnswer = ((random(0..14) + lastAnswer * 4) % 16) / 4
Set up a weighted array. Lets say the last value was a 2. Make an array like this:
array = [0,0,0,0,1,1,1,1,2,3,3,3,3];
Then pick a number in the array.
newValue = array[arc4random() % 13];
Now switch to using math instead of an array.
newValue = ( ( ( arc4random() % 13 ) / 4 ) + 1 + oldValue ) % 4;
For P possibilities and a weight 0<W<=1 use:
newValue = ( ( ( arc4random() % (P/W-P(1-W)) ) * W ) + 1 + oldValue ) % P;
For P=4 and W=1/4, (P/W-P(1-W)) = 13. This says the last value will be 1/4 as likely as other values.
If you completely eliminate the most recent answer it will be just as noticeable as the most recent answer showing up too often. I do not know what weight will feel right to you, but 1/4 is a good starting point.