Data Structure / Hash Function to link Sets of Ints to Value - data-structures

Given n integer id's, I wish to link all possible sets of up to k id's to a constant value. What I'm looking for is a way to translate sets (e.g. {1, 5}, {1, 3, 5} and {1, 2, 3, 4, 5, 6, 7}) to unique values.
Guarantees:
n < 100 and k < 10 (again: set sizes will range in [1, k]).
The order of id's doesn't matter: {1, 5} == {5, 1}.
All combinations are possible, but some may be excluded.
All sets and values are constant and made only once. No deletes or inserts, no value updates.
Once generated, the only operations taking place will be look-ups.
Look-ups will be frequent and one-directional (given set, look up value).
There is no need to sort (or otherwise organize) the values.
Additionally, it would be nice (but not obligatory) if "neighboring" sets (drop one id, add one id, swap one id, etc) are easy to reach, as well as "all sets that include at least this set".
Any ideas?

Enumerate using the product of primes.
a -> 2
b -> 3
c -> 5
d -> 7
et cetera
Now hash(ab) := 6, and hash (abc) := 30
And a nice side effect is that, if "ab" is a subset of "abc", then:
hash(abc) % hash(ab) == 0
and
hash(abc) / hash(ab) == hash(c)
The bad news: You might run into overflow, the 100th prime will probably be around 1000, and 64 bits cannot accomodate 1000**10. This will not affect the functioning as a hash function; only the subset thingy will fail to work. the same method applied to anagrams
The other option is Zobrist-hashing. It is equivalent to the the primes method, but instead of primes you use a fixed set of (random) numbers, and instead of multiplying you use XOR.
For a fixed small (it needs << ~70 bits) set like yours, it might be possible to tune the zobrist tables to totally avoid collisions (yielding a perfect hash).
And the final (and simplest) way is to use a (100bit) bitmap, and treat that as a hashvalue (maybe after modulo table size)
And a totally unrelated method is to just build a decision tree on the bits of the bitmap. (the tree would have a maximal depth of k) a related kD tree on bit values

May be not the best solution, but you can do the following:
Sort the set from Lowest to highest with a simple IntegerComparator
Add each item of the set to a String
so if you have {2,5,9,4} first Step->{2,4,5,9}; second->"2459"
This way you will get a unique String from a unique set. If you really need to map them to an integer value, you can hash the string after that.
A second way I can think of is to store them in a java Set and simply map it against a HashMap with set as keys

Calculate a 'diff' from each set {1, 6, 87, 89} = {1,5,81,2,0,0,...}
{1,2,3,4} = { 1,1,1,1,0,0,0,0... };
Then binary encode each number with a variable length encoding and concatenate the bits.
It's hard to compare the sets (except for the first few equal bits), but because there can't be many large intervals in a set, all possible values just might fit into 64 bits. (slack of 16 bits at least...)

Related

Preallocate or change size of vector

I have a situation where I have a process which needs to "burn-in". This means that I
Start with p values, p relatively small
For n>p, generate nth value using most recently generated p values (e.g. p+1 generated from values 1 to p, p+2 generated from values 2, p+1, etc.)
Repeat until n=N, where N large
Now, only the most recently generated p values will be useful to me, so there are two ways for me to implement this. I can either
Start with a vector of p initial values. At each iteration, mutate the vector, removing the first element, and replacing the last element with the most recently generated value or,
Preallocate a large array of length N, where first p elements are initial values. At iteration n, mutate nth value with most recently generated value
There are pros and cons to both approaches.
Pros of the first, are that we only store most relevant values. Cons of the first are that we are changing the length of the vector at each iteration.
Pros of the second are that we preallocate all the memory we need. Cons of the second is that we store much more than we need.
What is the best way to proceed? Does it depend on what aspect of performance I most need to care about? What will be the quickest?
Cheers in advance.
edit: approximately, p is usually in the order of low tens, N can be several thousand
The first solution has another huge cons: removing the first item of an array takes O(n) time since elements should be moved in memory. This certainly cause the algorithm to runs in quadratic time which is not reasonable. Shifting the items as proposed by #ForceBru should also cause this quadratic run time (since many items are moved just to add one value every time).
The second solution should be pretty fast compared to the first but, indeed, it can use a lot of memory so it should be sub-optimal (it takes time to write values in the RAM).
A faster solution is to use a data structure called a deque. Such data structure enable you to remove the first item in constant time and append a new value at the end also in constant time. That being said, it also introduces some overhead to be able to do that. Julia provide such data structure (more especially queues).
Since the number of in-flight items appears to be bounded in your algorithm, you can implement a rolling buffer. Fortunately, Julia also implement this: see CircularBuffer. This solution should be quite simple and fast (since the operations you want to do are done in O(1) time on it).
It is probably simplest to use CircularArrays.jl for your use case:
julia> using CircularArrays
julia> c = CircularArray([1,2,3,4])
4-element CircularVector(::Vector{Int64}):
1
2
3
4
julia> for i in 5:10
c[i] = i
#show c
end
c = [5, 2, 3, 4]
c = [5, 6, 3, 4]
c = [5, 6, 7, 4]
c = [5, 6, 7, 8]
c = [9, 6, 7, 8]
c = [9, 10, 7, 8]
In this way - as you can see - you can can continue using an increasing index and array will wrap around internally as needed (discarding old values that are not needed any more).
In this way you always store last p values in the array without having to copy anything or re-allocate memory in each step.
...only the most recently generated p values will be useful to me...
Start with a vector of p initial values. At each iteration, mutate the vector, removing the first element, and replacing the last element with the most recently generated value.
Cons of the first are that we are changing the length of the vector at each iteration.
There's no need to change the length of the vector. Simply shift its elements to the left (overwriting the first element) and write the new data to the_vector[end]:
the_vector = [1,2,3,4,5,6]
function shift_and_add!(vec::AbstractVector, value)
vec[1:end-1] .= #view vec[2:end] # shift
vec[end] = value # replace the last value
vec
end
#assert shift_and_add!(the_vector, 80) == [2,3,4,5,6,80]
# `the_vector` will be mutated
#assert the_vector == [2,3,4,5,6,80]

Sum of Function defined on Subsets

I want to know if their are any fast approaches to solve the following problem. I have a list of codes somewhere in the thousands (A0, A1, A2, ...). There is a positive value attached to about a million distinct combinations (A0-A1, A2-A10, A1-A2-A10, ...). Let the values be denoted f(A0-A1). Note that not all the combinations have the value attached.
For each listed combination, I want to calculate the sum of values of the values attached to each set that contains the given combination. For instance, for A2-A10,
calculate
g(A2-A10) = f(A2-A10) + f(A1-A2-A10) + ...
I would like to do this with minimal time complexity. A simpler related problem is to find all combinations where g(C) is greater than a threshold value.
Key the existing combinations with a bit map, where bit n denotes whether An is in that particular coding. Store the values keyed by the bit map for each in your favorite hash-map structure. Thus, f(A0, A1, A10, A12) would be combo_val[11000000001010000...]
To sum all of the desired combinations, build a bit map of your root. For instance, with the combination above, we'd have root = 1100000000101000 (cutting off at 16 total elements for the sake of illustration.
Now simply loop through the keys of the hashmap, using root as a mask. Sum the desired values:
total = 0
for key in combo_val.keys()
if root && key == root
total += combo_val[key]
Does that get you moving?
I thought waaay too long before coming up with the following approach.
Index the million combinations. So you know which you want. In your example:
0: A0-A1
1: A2-A10
2: A1-A2-A10
For each code, create an ordered list of combinations that contain that code. Call that code_combs. In your example:
A0: [0]
A1: [0, 2]
A2: [1, 2]
A10: [1, 2]
Now we have a combination of codes, like A2-A10. We create two arrays, one of codes, the other of indices. Set indices at 0. So:
codes = ['A2', 'A10']
indices = [0, 0]
And now do the following:
while not done:
let max_comb = max(code_combs[codes[i]][indices[i]] over i in range(len(codes))
Advance each index until we are at the max_comb or greater
(if we reach the end of any list, we are done)
If all are at the same max_comb, we add its value.
Advance all indexes by 1.
(if we reach the end of any list, we are done)
Basically this is a k-way intersection of ordered lists. Now here is the trick. If we advance naively, this will be slightly faster because we only have to look at combinations that contain a code. However we can use a clever advance strategy like this:
Advance by 1, 2, 4, 8, etc until we reach or pass the point we want.
Do a binary search between the last two values until we find the point we want
(Be warned, implementing binary search is not always so easy to get right.)
And now we are crossing fingers. But if any one of our codes has few combinations that it is in, and there aren't too many codes in our combination, we can compute our intersection quite quickly.

Compare rotated lists, containing duplicates [duplicate]

This question already has answers here:
How to check whether two lists are circularly identical in Python
(18 answers)
Closed 7 years ago.
I'm looking for an efficient way to compare lists of numbers to see if they match at any rotation (comparing 2 circular lists).
When the lists don't have duplicates, picking smallest/largest value and rotating both lists before comparisons works.
But when there may be many duplicate large values, this isn't so simple.
For example, lists [9, 2, 0, 0, 9] and [0, 0, 9, 9, 2] are matches,where [9, 0, 2, 0, 9] won't (since the order is different).
Heres an example of an in-efficient function which works.
def min_list_rotation(ls):
return min((ls[i:] + ls[:i] for i in range(len(ls))))
# example use
ls_a = [9, 2, 0, 0, 9]
ls_b = [0, 0, 9, 9, 2]
print(min_list_rotation(ls_a) == min_list_rotation(ls_b))
This can be improved on for efficiency...
check sorted lists match before running exhaustive tests.
only test rotations that start with the minimum value(skipping matching values after that)effectively finding the minimum value with the furthest & smallest number after it (continually - in the case there are multiple matching next-biggest values).
compare rotations without creating the new lists each time..
However its still not a very efficient method since it relies on checking many possibilities.
Is there a more efficient way to perform this comparison?
Related question:
Compare rotated lists in python
If you are looking for duplicates in a large number of lists, you could rotate each list to its lexicographically minimal string representation, then sort the list of lists or use a hash table to find duplicates. This canonicalisation step means that you don't need to compare every list with every other list. There are clever O(n) algorithms for finding the minimal rotation described at https://en.wikipedia.org/wiki/Lexicographically_minimal_string_rotation.
You almost have it.
You can do some kind of "normalization" or "canonicalisation" of a list independently of the others, then you only need to compare item by item (or if you want, put them in a map, in a set to eliminate duplicates, ..."
1 take the minimum item, which is not preceded by itself (in a circular way)
In you example 92009, you should take the first 0 (not the second one)
2 If you have always the same item (say 00000), you just keep that: 00000
3 If you have the same item several times, take the next item, which is minimal, and keep going until you find one unique path with minimums.
Example: 90148301562 => you have 0148.. and 0156.. => you take 0148
4 If you can not separate the different paths (= if you have equality at infinite), you have a repeating pattern: then, no matters: you take any of them.
Example: 014376501437650143765 : you have the same pattern 0143765...
It is like AAA, where A = 0143765
5 When you have your list in this form, it is easy to compare two of them.
How to do that efficiently:
Iterate on your list to get the minimums Mx (not preceded by itself). If you find several, keep all of them.
Then, iterate from each minimum Mx, take the next item, and keep the minimums. If you do an entire cycle, you have a repeating pattern.
Except the case of repeating pattern, this must be the minimal way.
Hope it helps.
I would do this in expected O(N) time using a polynomial hash function to compute the hash of list A, and every cyclic shift of list B. Where a shift of list B has the same hash as list A, I'd compare the actual elements to see if they are equal.
The reason this is fast is that with polynomial hash functions (which are extremely common!), you can calculate the hash of each cyclic shift from the previous one in constant time, so you can calculate hashes for all of the cyclic shifts in O(N) time.
It works like this:
Let's say B has N elements, then the the hash of B using prime P is:
Hb=0;
for (i=0; i<N ; i++)
{
Hb = Hb*P + B[i];
}
This is an optimized way to evaluate a polynomial in P, and is equivalent to:
Hb=0;
for (i=0; i<N ; i++)
{
Hb += B[i] * P^(N-1-i); //^ is exponentiation, not XOR
}
Notice how every B[i] is multiplied by P^(N-1-i). If we shift B to the left by 1, then every every B[i] will be multiplied by an extra P, except the first one. Since multiplication distributes over addition, we can multiply all the components at once just by multiplying the whole hash, and then fix up the factor for the first element.
The hash of the left shift of B is just
Hb1 = Hb*P + B[0]*(1-(P^N))
The second left shift:
Hb2 = Hb1*P + B[1]*(1-(P^N))
and so on...

Subset counting algorithm

I have a following problem I want to solve efficiently. I am given a set of k-tuples of Boolean values where I know in advance that some fraction of each of the values in each of the k-tuples is true. For example, I might have the following 4-tuples, where each tuple has at least 60% of it's Boolean values set to true:
(1, 0, 1, 0)
(1, 1, 0, 1)
(0, 0, 1, 0)
I am interested in finding sets of indices that have a particular property: if I look at each of the values in the tuples at the indicated indices, at least the given fraction of those tuples have the corresponding bit set. For example, in the above set of 4-tuples, I could consider the set {0}, since if you look at the zeroth element of each of the above tuples, two-thirds of them are 1, and 2/3 ~= 66% > 60%. I could also consider the set {2} for the same reason. However, I could not consider {1}, since at index 1 only one third of the tuples have a 1 and 1/3 is less than 60%. Similarly, I could not use {0, 2} as a set, because it is not true that at least 60% of the tuples have both bits 0 and 2 set.
My goal is to find all sets for which this property holds. Does anyone have a good algorithm for solving this?
Thank you.
As you've wrote, that can be assumed that architecture is x86_64 and you are looking for implementation performance, cause asymptotic complexity (as it is not going to go under linear - by definition of problem ;) ), I propose following algorithm (C++ like pseudocode):
/* N=16 -> int16; N=8 -> int8 etc. Select N according to input sizes. (maybe N=24 ;) ) */
count_occurences_intN(vector<intN> t, vector<long> &result_counters){
intN counters[2^N]={};
//first, count bit combinations
for_each(v in t)
++counters[v];
//second, count bit occurrences, using aggregated data
for(column=0; column<N; ++column){
mask = 1 << column;
long *result_counter_ptr = &(result_counters[column]);
for(v=0; v<2^16; ++v)
if( v & mask )
++(*result_counter_ptr);
}
}
Than, split your input k-bit vectors into N-bit vectors, and apply above function.
Depending on input size you might improve performance you choosing N=8, N=16, N=24 or applying naive approach.
As you've wrote, you can not assume anything on client side, just implement N={8,16,24} and naive and select one from four implementations depending on size of input.
Make a k-vector of integers, describing how many passes there were for each index. Loop through your set, for each element incrementing the k-vector of passes.
Then figure out the cardinality of your set (either in a separate loop, or in the above one). Then loop through your vector of counts, and emit a pass/fail vector based on your criteria.

sorting algorithm where pairwise-comparison can return more information than -1, 0, +1

Most sort algorithms rely on a pairwise-comparison the determines whether A < B, A = B or A > B.
I'm looking for algorithms (and for bonus points, code in Python) that take advantage of a pairwise-comparison function that can distinguish a lot less from a little less or a lot more from a little more. So perhaps instead of returning {-1, 0, 1} the comparison function returns {-2, -1, 0, 1, 2} or {-5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5} or even a real number on the interval (-1, 1).
For some applications (such as near sorting or approximate sorting) this would enable a reasonable sort to be determined with less comparisons.
The extra information can indeed be used to minimize the total number of comparisons. Calls to the super_comparison function can be used to make deductions equivalent to a great number of calls to a regular comparsion function. For example, a much-less-than b and c little-less-than b implies a < c < b.
The deductions cans be organized into bins or partitions which can each be sorted separately. Effectively, this is equivalent to QuickSort with n-way partition. Here's an implementation in Python:
from collections import defaultdict
from random import choice
def quicksort(seq, compare):
'Stable in-place sort using a 3-or-more-way comparison function'
# Make an n-way partition on a random pivot value
segments = defaultdict(list)
pivot = choice(seq)
for x in seq:
ranking = 0 if x is pivot else compare(x, pivot)
segments[ranking].append(x)
seq.clear()
# Recursively sort each segment and store it in the sequence
for ranking, segment in sorted(segments.items()):
if ranking and len(segment) > 1:
quicksort(segment, compare)
seq += segment
if __name__ == '__main__':
from random import randrange
from math import log10
def super_compare(a, b):
'Compare with extra logarithmic near/far information'
c = -1 if a < b else 1 if a > b else 0
return c * (int(log10(max(abs(a - b), 1.0))) + 1)
n = 10000
data = [randrange(4*n) for i in range(n)]
goal = sorted(data)
quicksort(data, super_compare)
print(data == goal)
By instrumenting this code with the trace module, it is possible to measure the performance gain. In the above code, a regular three-way compare uses 133,000 comparisons while a super comparison function reduces the number of calls to 85,000.
The code also makes it easy to experiment with a variety comparison functions. This will show that naïve n-way comparison functions do very little to help the sort. For example, if the comparison function returns +/-2 for differences greater than four and +/-1 for differences four or less, there is only a modest 5% reduction in the number of comparisons. The root cause is that the course grained partitions used in the beginning only have a handful of "near matches" and everything else falls in "far matches".
An improvement to the super comparison is to covers logarithmic ranges (i.e. +/-1 if within ten, +/-2 if within a hundred, +/- if within a thousand.
An ideal comparison function would be adaptive. For any given sequence size, the comparison function should strive to subdivide the sequence into partitions of roughly equal size. Information theory tells us that this will maximize the number of bits of information per comparison.
The adaptive approach makes good intuitive sense as well. People should first be partitioned into love vs like before making more refined distinctions such as love-a-lot vs love-a-little. Further partitioning passes should each make finer and finer distinctions.
You can use a modified quick sort. Let me explain on an example when you comparison function returns [-2, -1, 0, 1, 2]. Say, you have an array A to sort.
Create 5 empty arrays - Aminus2, Aminus1, A0, Aplus1, Aplus2.
Pick an arbitrary element of A, X.
For each element of the array, compare it with X.
Depending on the result, place the element in one of the Aminus2, Aminus1, A0, Aplus1, Aplus2 arrays.
Apply the same sort recursively to Aminus2, Aminus1, Aplus1, Aplus2 (note: you don't need to sort A0, as all he elements there are equal X).
Concatenate the arrays to get the final result: A = Aminus2 + Aminus1 + A0 + Aplus1 + Aplus2.
It seems like using raindog's modified quicksort would let you stream out results sooner and perhaps page into them faster.
Maybe those features are already available from a carefully-controlled qsort operation? I haven't thought much about it.
This also sounds kind of like radix sort except instead of looking at each digit (or other kind of bucket rule), you're making up buckets from the rich comparisons. I have a hard time thinking of a case where rich comparisons are available but digits (or something like them) aren't.
I can't think of any situation in which this would be really useful. Even if I could, I suspect the added CPU cycles needed to sort fuzzy values would be more than those "extra comparisons" you allude to. But I'll still offer a suggestion.
Consider this possibility (all strings use the 27 characters a-z and _):
11111111112
12345678901234567890
1/ now_is_the_time
2/ now_is_never
3/ now_we_have_to_go
4/ aaa
5/ ___
Obviously strings 1 and 2 are more similar that 1 and 3 and much more similar than 1 and 4.
One approach is to scale the difference value for each identical character position and use the first different character to set the last position.
Putting aside signs for the moment, comparing string 1 with 2, the differ in position 8 by 'n' - 't'. That's a difference of 6. In order to turn that into a single digit 1-9, we use the formula:
digit = ceiling(9 * abs(diff) / 27)
since the maximum difference is 26. The minimum difference of 1 becomes the digit 1. The maximum difference of 26 becomes the digit 9. Our difference of 6 becomes 3.
And because the difference is in position 8, out comparison function will return 3x10-8 (actually it will return the negative of that since string 1 comes after string 2.
Using a similar process for strings 1 and 4, the comparison function returns -5x10-1. The highest possible return (strings 4 and 5) has a difference in position 1 of '-' - 'a' (26) which generates the digit 9 and hence gives us 9x10-1.
Take these suggestions and use them as you see fit. I'd be interested in knowing how your fuzzy comparison code ends up working out.
Considering you are looking to order a number of items based on human comparison you might want to approach this problem like a sports tournament. You might allow each human vote to increase the score of the winner by 3 and decrease the looser by 3, +2 and -2, +1 and -1 or just 0 0 for a draw.
Then you just do a regular sort based on the scores.
Another alternative would be a single or double elimination tournament structure.
You can use two comparisons, to achieve this. Multiply the more important comparison by 2, and add them together.
Here is a example of what I mean in Perl.
It compares two array references by the first element, then by the second element.
use strict;
use warnings;
use 5.010;
my #array = (
[a => 2],
[b => 1],
[a => 1],
[c => 0]
);
say "$_->[0] => $_->[1]" for sort {
($a->[0] cmp $b->[0]) * 2 +
($a->[1] <=> $b->[1]);
} #array;
a => 1
a => 2
b => 1
c => 0
You could extend this to any number of comparisons very easily.
Perhaps there's a good reason to do this but I don't think it beats the alternatives for any given situation and certainly isn't good for general cases. The reason? Unless you know something about the domain of the input data and about the distribution of values you can't really improve over, say, quicksort. And if you do know those things, there are often ways that would be much more effective.
Anti-example: suppose your comparison returns a value of "huge difference" for numbers differing by more than 1000, and that the input is {0, 10000, 20000, 30000, ...}
Anti-example: same as above but with input {0, 10000, 10001, 10002, 20000, 20001, ...}
But, you say, I know my inputs don't look like that! Well, in that case tell us what your inputs really look like, in detail. Then someone might be able to really help.
For instance, once I needed to sort historical data. The data was kept sorted. When new data were added it was appended, then the list was run again. I did not have the information of where the new data was appended. I designed a hybrid sort for this situation that handily beat qsort and others by picking a sort that was quick on already sorted data and tweaking it to be fast (essentially switching to qsort) when it encountered unsorted data.
The only way you're going to improve over the general purpose sorts is to know your data. And if you want answers you're going to have to communicate that here very well.

Resources