Matrix Math - Maximizing - python

I have a dataframe that with an index of magic card names. The columns are the same index, resulting in a 1081 x 1081 dataframe of each card in my collection paired with each other card in my collection.
I have code that identifies combos of cards that go well together. For example "Whenever you draw a card" pairs well with "Draw a card" cards. I find the junction of those two cards and increase its value by 1.
Now, I need to find the maximum value for 36 cards.
But, how?
Randomly selecting cards is useless, there are 1.717391336 E+74 potential combinations. I've tried pulling out the lowest values and that reduces the set of potential combinations, but even at 100 cards you're talking about 1.977204582 E+27 potentials.
This has to have been solved by someone smarter than me - can ya'll point me in the right direction?

As you pointed out already, the combinatorics are not on your side here. There are 1081 choose 36 possible sets (binomial coefficient), so it is out of question to check all of them.
I am not aware of any practicable solution to find the optimal set for the general problem, that is without knowing the 1081x1081 matrix.
For an approximate solution for the general problem, you might want to try a greedy approach, while keeping a history of n sets after each step, with e.g. n = 1000.
So you would start with going through all sets with 2 cards, which is 1081 * 1080 / 2 combinations, look up the value in the matrix for each and pick the n max ones.
In the second step, for each of the n kept sets, go through all possible combinations with a third card (and check for duplicate sets), i.e. checking n * 1079 sets, and keep the n max ones.
In the third step, check n * 1078 sets with a fourth card, and so on, and so forth.
Of course, this won't give you the optimal solution for the general case, but maybe it's good enough for your given situation. You can also take a look at the history, to get a feeling for how often it happens that the best set from step x is caught up by another set in a step y > x. Depending on your matrix, it might not happen that often or even never.

Related

Possibilities to Split Book-Stack

I'm struggeling with the following problem in Python. The problem is more like math specific than Python specific.
I've got a number of N books as a Stack. And a number of P stacks to split on.
I'm looking for the possibilites to split this stack, avoiding repetitions and empty stacks.
So let's say my stack is 4 books tall, what are the possibilities to split on 2 stacks ?
The possibilities to split would be:
(1,3)
(2,2)
There would also be the possibility of (3,1), but since (1,3) is already in my output, I don't want (3,1) to be there too.
Another example:
5 books, 3 stacks
(3,1,1)
(2,2,1)
Solutions like (1,1,3), (2,1,2) are not in my output beacause they are redundant.
Im looking for an EFFICIENT way to compute the tuples of stacks.
I'm working with a starting stack sizes up to 400, this stack could be split into another stack, which could also be split and so on.
Is there already a reference which covers this problem?
I think it would be easy to solve in combinatoric terms, but the problem here is, I am interested in the Possibilities themself and not just the number of possibilities !
Any help here?
cheers
Eliminating duplicates:
You can do this by taking the first permutation of each combination.
In other words having the smallest stacks in front.
E.g {1,2,3},{1,3,2},{2,1,3},{2,3,1},{3,1,2},{3,2,1}
Efficiency:
You probably want to do this with recursion, so at each step you know the possible size of the stack is at least the size of the previous
You know that all following stackSizes have to be at least the current size. So the maximum size is the number of books left divided by the number of stacks left (floor).
E.g. 10 books left for 3 stacks. floor(10/3) = 3. Which is right because the max combination left at that point is {3,3,4}
Hence this will prevent you to step into a failing combination.
Code
import math
def bookStack(cur, min, booksLeft, sizes):
if len(sizes) == (cur+1):
sizes[cur] = booksLeft
print(sizes)
return;
max = math.floor(booksLeft / (len(sizes)-cur))+1;
for take in range(min,max):
sizes[cur] = take
bookStack(cur+1, take, booksLeft-take, sizes)
For 5 books over 3 stacks, call this with:
bookStack(0,1,5,[0]*3)
Run it here
Remark: Although you want all unique combinations, this still is a fast growing function and will only work for a small number of stacks. Or when the number of stacks is almost equal with the number of books. You will notice.

Statistics: Optimizing probability calculations within python

Setup:
The question is complex form of a classic probability question:
70 colored balls are placed in an urn, 10 for each of the seven rainbow colors.
What is the expected number of distinct colors in 20 randomly picked balls?
My solution is python's itertools library:
combos = itertools.combinations(urn, 20),
print sum([1 for x in combos])
(where urn is a list of the 70 balls in the urn).
I can unpack the iterator up to a length of combinations(urn, 8) past that my computer can't handle it.
Note: I know this wouldn't give me the answer, this is only the road block in my script, in other words if this worked my script would work.
Question: How could I find the expected colors accurately, without the worlds fastest super computer? Is my way even computationally possible?
Since a couple of people have asked to see the mathematical solution, I'll give it. This is one of the Project Euler problems that can be done in a reasonable amount of time with pencil and paper. The answer is
7(1 - (60 choose 20)/(70 choose 20))
To get this write X, the count of colors present, as a sum X0+X1+X2+...+X6, where Xi is 1 if the ith color is present, and 0 if it is not present.
E(X)
= E(X0+X1+...+X6)
= E(X0) + E(X1) + ... + E(X6) by linearity of expectation
= 7E(X0) by symmetry
= 7 * probability that a particular color is present
= 7 * (1- probability that a particular color is absent)
= 7 * (1 - (# ways to pick 20 avoiding a color)/(# ways to pick 20))
= 7 * (1 - (60 choose 20)/(70 choose 20))
Expectation is always linear. So, when you are asked to find the average value of some random quantity, it often helps to try to rewrite the quantity as a sum of simpler pieces such as indicator (0-1) random variables.
This does not say how to make the OP's approach work. Although there is a direct mathematical solution, it is good to know how to iterate through the cases in an organized and practicable fashion. This could help if you next wanted a more complicated function of the set of colors present than the count. Duffymo's answer suggested something that I'll make more explicit:
You can break up the ways to draw 20 calls from 70 into categories indexed by the counts of colors. For example, the index (5,5,10,0,0,0,0) means we drew 5 of the first color, 5 of the second color, 10 of the third color, and none of the other colors.
The set of possible indices is contained in the collection of 7-tuples of nonnegative integers with sum 20. Some of these are impossible, such as (11,9,0,0,0,0,0) by the problem's assumption that there are only 10 balls of each color, but we can deal with that. The set of 7-tuples of nonnegative numbers adding up to 20 has size (26 choose 6)=230230, and it has a natural correspondence with the ways of choosing 6 dividers among 26 spaces for dividers or objects. So, if you have a way to iterate through the 6 element subsets of a 26 element set, you can convert these to iterate through all indices.
You still have to weight the cases by the counts of the ways to draw 20 balls from 70 to get that case. The weight of (a0,a1,a2,...,a6) is (10 choose a0)(10 choose a1)...*(10 choose a6). This handles the case of impossible indices gracefully, since 10 choose 11 is 0 so the product is 0.
So, if you didn't know about the mathematical solution by the linearity of expectation, you could iterate through 230230 cases and compute a weighted average of the number of nonzero coordinates of the index vector, weighted by a product of small binomial terms.
Wouldn't it just be combinations with repetition?
http://www.mathsisfun.com/combinatorics/combinations-permutations.html
Make an urn with 10 of each color.
Decide on the number of trials you want.
Make a container to hold the result of each trial
for each trial, pick a random sample of twenty items from the urn, make a set of those items, add the length of that set to the results.
find the average of the results

Pebbling a Checkerboard with Dynamic Programming

I am trying to teach myself Dynamic Programming, and ran into this problem from MIT.
We are given a checkerboard which has 4 rows and n columns, and
has an integer written in each square. We are also given a set of 2n pebbles, and we want to
place some or all of these on the checkerboard (each pebble can be placed on exactly one square)
so as to maximize the sum of the integers in the squares that are covered by pebbles. There is
one constraint: for a placement of pebbles to be legal, no two of them can be on horizontally or
vertically adjacent squares (diagonal adjacency is ok).
(a) Determine the number of legal patterns that can occur in any column (in isolation, ignoring
the pebbles in adjacent columns) and describe these patterns.
Call two patterns compatible if they can be placed on adjacent columns to form a legal placement.
Let us consider subproblems consisting of the rst k columns 1 k n. Each subproblem can
be assigned a type, which is the pattern occurring in the last column.
(b) Using the notions of compatibility and type, give an O(n)-time dynamic programming algorithm for computing an optimal placement.
Ok, so for part a: There are 8 possible solutions.
For part b, I'm unsure, but this is where I'm headed:
SPlit into sub-problems. Assume i in n.
1. Define Cj[i] to be the optimal value by pebbling columns 0,...,i, such that column i has pattern type j.
2. Create 8 separate arrays of n elements for each pattern type.
I am not sure where to go from here. I realize there are solutions to this problem online, but the solutions don't seem very clear to me.
You're on the right track. As you examine each new column, you will end up computing all possible best-scores up to that point.
Let's say you built your compatibility list (a 2D array) and called it Li[y] such that for each pattern i there are one or more compatible patterns Li[y].
Now, you examine column j. First, you compute that column's isolated scores for each pattern i. Call it Sj[i]. For each pattern i and compatible
pattern x = Li[y], you need to maximize the total score Cj such that Cj[x] = Cj-1[i] + Sj[x]. This is a simple array test and update (if bigger).
In addition, you store the pebbling pattern that led to each score. When you update Cj[x] (ie you increase its score from its present value) then remember the initial and subsequent patterns that caused the update as Pj[x] = i. That says "pattern x gave the best result, given the preceding pattern i".
When you are all done, just find the pattern i with the best score Cn[i]. You can then backtrack using Pj to recover the pebbling pattern from each column that led to this result.

How to generate statistically probably locations for ships in battleship

I made the original battleship and now I'm looking to upgrade my AI from random guessing to guessing statistically probably locations. I'm having trouble finding algorithms online, so my question is what kinds of algorithms already exist for this application? And how would I implement one?
Ships: 5, 4, 3, 3, 2
Field: 10X10
Board:
OCEAN = "O"
FIRE = "X"
HIT = "*"
SIZE = 10
SEA = [] # Blank Board
for x in range(SIZE):
SEA.append([OCEAN] * SIZE)
If you'd like to see the rest of the code, I posted it here: (https://github.com/Dbz/Battleship/blob/master/BattleShip.py); I didn't want to clutter the question with a lot of irrelevant code.
The ultimate naive solution wold be to go through every possible placement of ships (legal given what information is known) and counting the number of times each square is full.
obviously, in a relatively empty board this will not work as there are too many permutations, but a good start might be:
for each square on board: go through all ships and count in how many different ways it fits in that square, i.e. for each square of the ships length check if it fits horizontally and vertically.
an improvement might be to also check for each possible ship placement if the rest of the ships can be placed legally whilst covering all known 'hits' (places known to contain a ship).
to improve performance, if only one ship can be placed in a given spot, you no longer need to test it on other spots. also, when there are many 'hits', it might be quicker to first cover all known 'hits' and for each possible cover go through the rest.
edit: you might want to look into DFS.
Edit: Elaboration on OP's (#Dbz) suggestion in the comments:
hold a set of dismissed placements ('dissmissed') of ships (can be represented as string, say "4V5x3" for the placement of length 4 ship in 5x3, 5x4, 5x5, 5x6), after a guess you add all the placements the guess dismisses, then for each square hold a set of placements that intersect with it ('placements[x,y]') then the probability would be:
34-|intersection(placements[x,y], dissmissed)|/(3400-|dismissed|)
To add to the dismissed list:
if guess at (X,Y) is a miss add placements[x,y]
if guess at (X,Y) is a hit:
add neighboring placements (assuming that ships cannot be placed adjacently), i.e. add:
<(2,3a,3b,4,5)>H<X+1>x<Y>, <(2,3a,3b,4,5)>V<X>x<Y+1>
<(2,3a,3b,4,5)>H<X-(2,3,3,4,5)>x<Y>, <(2,3a,3b,4,5)>V<X>x<Y-(2,3,3,4,5)>
2H<X+-1>x<Y+(-2 to 1)>, 3aH<X+-1>x<Y+(-3 to 1)> ...
2V<X+(-2 to 1)>x<Y+-1>, 3aV<X+(-3 to 1)>x<Y+-1> ...
if |intersection(placements[x,y], dissmissed)|==33, i.e. only one placement possible add ship (see later)
check if any of the previews hits has only one possible placement left, if so, add the ship
check to see if any of the ships have only possible placement, if so, add the ship
adding a ship:
add all other placements of that ship to dismissed
for each (x,y) of the ships placement add placements[x,y] with out the actual placement
for each (x,y) of the ships placement mark as hit guess (if not already known) run stage 2
for each (x,y) neighboring the ships placement mark as miss guess (if not already known) run stage 1
run stage 3 and 4.
i might have over complicated this, there might be some redundant actions, but you get the point.
Nice question, and I like your idea for statistical approach.
I think I would have tried a machine learning approach for this problem as follows:
First model your problem as a classification problem.
The classification problem is: Given a square (x,y) - you want to tell the likelihood of having a ship in this square. Let this likelihood be p.
Next, you need to develop some 'features'. You can take the surrounding of (x,y) [as you might have partial knowledge on it] as your features.
For example, the features of the middle of the following mini-board (+ indicates the square you want to determine if there is a ship or not in):
OO*
O+*
?O?
can be something like:
f1 = (0,0) = false
f2 = (0,1) = false
f3 = (0,2) = true
f4 = (1,0) = false
**note skipping (1,1)
f5 = (1,2) = true
f6 = (2,0) = unknown
f7 = (2,1) = false
f8 = (2,2) = unknown
I'd implement features relative to the point of origin (in this case - (1,1)) and not as absolute location on board (so the square up to (3,3) will also be f2).
Now, create a training set. The training set is a 'labeled' set of features - based on some real boards. You can create it manually (create a lot of boards), automatically by a random generator of placements, or by some other data you can gather.
Feed the training set to a learning algorithm. The algorithm should be able to handle 'unknowns' and be able to give probability of "true" and not only a boolean answer. I think a variation of Naive Bayes can fit well here.
After you have got a classifier - exploit it with your AI.
When it's your turn, choose to fire upon a square which has the maximal value of p. At first, the shots will be kinda random - but with more shots you fire, you will have more information on the board, and the AI will exploit it for better predictions.
Note that I gave features based on a square of size 1. You can of course choose any k and find features on this bigger square - it will give you more features, but each might be less informative. There is no rule of thumb which will be better - and it should be tested.
Main question is, how are you going to find statistically probable locations. Are they already known or you want to figure them out?
Either case, I'd just make the grid weighed. In your case, the initial weight for each slot would be 1.0/(SIZE^2). The sum of weights must be equal to 1.
You can then adjust weights based on the statistics gathered from N last played games.
Now, when your AI makes a choice, it chooses a coordinate to hit based on weighed probabilities. The quick and simple way to do that would be:
Generate a random number R in range [0..1]
Start from slot (0, 0) adding the weights, i.e. S = W(0, 0) + W(0, 1) + .... where W(n, m) is the weight of the corresponding slot. Once S >= R, you've got the coordinate to hit.
This can be optimised by pre-calculating cumulative weights for each row, have fun :)
Find out which ships are still alive:
alive = [2,2,3,4] # length of alive ships
Find out spots where you have not shot, for example with a numpy.where()
Loop over spots where you can shoot.
Check the sides of the given position. Go left and right, how many spaces? Go up and down, how many spaces? If you can fit a boat in that many spaces, you can fit any smaller boat, so this loop I'd do it from the largest ship downwards, and I'd add to the counts in this position as many +1 as ships smaller than the one that fits.
Once you have done all of this, the position with more points should be the most probable to attack and hit something.
Of course, it can get as complicated as you want. You can also ask yourself, instead of which is my next hit, which combinations of hits will give me the victory in less number of hits or any other combination/parametrization of the problem. Good luck!

Challenging dynamic programming problem

This is a toned down version of a computer vision problem I need to solve. Suppose you are given parameters n,q and have to count the number of ways of assigning integers 0..(q-1) to elements of n-by-n grid so that for each assignment the following are all true
No two neighbors (horizontally or vertically) get the same value.
Value at positions (i,j) is 0
Value at position (k,l) is 0
Since (i,j,k,l) are not given, the output should be an array of evaluations above, one for every valid setting of (i,j,k,l)
A brute force approach is below. The goal is to get an efficient algorithm that works for q<=100 and for n<=18.
def tuples(n,q):
return [[a,]+b for a in range(q) for b in tuples(n-1,q)] if n>1 else [[a] for a in range(q)]
def isvalid(t,n):
grid=[t[n*i:n*(i+1)] for i in range(n)];
for r in range(n):
for c in range(n):
v=grid[r][c]
left=grid[r][c-1] if c>0 else -1
right=grid[r][c-1] if c<n-1 else -1
top=grid[r-1][c] if r > 0 else -1
bottom=grid[r+1][c] if r < n-1 else -1
if v==left or v==right or v==top or v==bottom:
return False
return True
def count(n,q):
result=[]
for pos1 in range(n**2):
for pos2 in range(n**2):
total=0
for t in tuples(n**2,q):
if t[pos1]==0 and t[pos2]==0 and isvalid(t,n):
total+=1
result.append(total)
return result
assert count(2,2)==[1, 0, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1]
Update 11/11
I've also asked this on TopCoder forums, and their solution is the most efficient one I've seen so far (about 3 hours for n=10, any q, from author's estimate)
Maybe this sounds too simple, but it works. Randomly distribute values to all the cells until only two are empty. Test for adjacency of all values. Compute the average the percent of successful casts vs. all casts until the variance drops to within an acceptable margin.
The risk goes to zero and the that which is at risk is only a little runtime.
This isn't an answer, just a contribution to the discussion which is too long for a comment.
tl; dr; Any algorithm which boils down to, "Compute the possibilities and count them," such as Eric Lippert's or a brute force approach won't work for #Yaroslav's goal of q <= 100 and n <= 18.
Let's first think about a single n x 1 column. How many valid numberings of this one column exist? For the first cell we can pick between q numbers. Since we can't repeat vertically, we can pick between q - 1 numbers for the second cell, and therefore q - 1 numbers for the third cell, and so on. For q == 100 and n == 18 that means there are q * (q - 1) ^ (n - 1) = 100 * 99 ^ 17 valid colorings which is very roughly 10 ^ 36.
Now consider any two valid columns (call them the bread columns) separated by a buffer column (call it the mustard column). Here is a trivial algorithm to find a valid set of values for the mustard column when q >= 4. Start at the top cell of the mustard column. We only have to worry about the adjacent cells of the bread columns which have at most 2 unique values. Pick any third number for the mustard column. Consider the second cell of the mustard column. We must consider the previous mustard cell and the 2 adjacent bread cells with a total of at most 3 unique values. Pick the 4th value. Continue to fill out the mustard column.
We have at most 2 columns containing a hard coded cell of 0. Using mustard columns, we can therefore make at least 6 bread columns, each with about 10 ^ 36 solutions for a total of at least 10 ^ 216 valid solutions, give or take an order of magnitude for rounding errors.
There are, according to Wikipedia, about 10 ^ 80 atoms in the universe.
Therefore, be cleverer.
Update 11/11 I've also asked this on TopCoder forums, and their solution is the most efficient one I've seen so far (about 41 hours hours for n=10, any q, from author's estimate)
I'm the author. Not 41, just 3 embarrassingly parallelizable CPU hours. I've counted symmetries. For n=10 there are only 675 really distinct pairs of (i,j) and (k,l). My program needs ~ 16 seconds per each.
I'm building a contribution based on the contribution to the discussion by Dave Aaron Smith.
Let's not consider for now the last two constraints ((i,j) and (k,l)).
With only one column (nx1) the solution is q * (q - 1) ^ (n - 1).
How many choices for a second column ? (q-1) for the top cell (1,2) but then q-1 or q-2 for the cell (2,2) if (1,2)/(2,1) have or not the same color.
Same thing for (3,2) : q-1 or q-2 solutions.
We can see we have a binary tree of possibilities and we need to sum over that tree. Let's assume left child is always "same color on top and at left" and right child is "different colors".
By computing over the tree the number of possibilities for the left column to create a such configurations and the number of possibilities for the new cells we are coloring we would count the number of possibilities for coloring two columns.
But let's now consider the probability distribution foe the coloring of the second column : if we want to iterate the process, we need to have an uniform distribution on the second column, it would be like the first one never existed and among all coloring of the first two column we could say things like 1/q of them have color 0 in the top cell of second column.
Without an uniform distribution it would be impossible.
The problem : is the distribution uniform ?
Answer :
We would have obtain the same number of solution by building first the second column them the first one and then the third one. The distribution of the second column is uniform in that case so it also is in the first case.
We can now apply the same "tree idea" to count the number of possibilities for the third column.
I will try to develop on that and build a general formula (since the tree is of size 2^n we don't want to explicitly explore it).
A few observations which might help other answerers as well:
The values 1..q are interchangeable - they could be letters and the result would be the same.
The constraints that no neighbours match is a very mild one, so a brute force approach will be excessively expensive. Even if you knew the values in all but one cell, there would still be at least q-8 possibilities for q>8.
The output of this will be pretty long - every set of i,j,k,l will need a line. The number of combinations is something like n2(n2-3), since the two fixed zeroes can be anywhere except adjacent to each other, unless they need not obey the first rule. For n=100 and q=18, the maximally hard case, this is ~ 1004 = 100 million. So that's your minimum complexity, and is unavoidable as the problem is currently stated.
There are simple cases - when q=2, there are the two possible checkerboards, so for any given pair of zeroes the answer is 1.
Point 3 makes the whole program O( n2(n2-3) ) as a minimum, and also suggests that you will need something reasonably efficient for each pair of zeroes as simply writing 100 million lines without any computation will take a while. For reference, at a second per line, that is 1x108s ~ 3 years, or 3 months on a 12-core box.
I suspect that there is an elegant answer given a pair of zeroes, but I'm not sure that there is an analytic solution to it. Given that you can do it with 2 or 3 colours depending on the positions of the zeroes, you could split the map into a series of regions, each of which uses only 2 or 3 colours, and then it's just the number of different combinations of 2 or 3 in q (qC2 or qC3) for each region times the number of regions, times the number of ways of splitting the map.
I'm not a mathematician, but it occurs to me that there ought to be an analytical solution to this problem, namely:
First, compute now many different colourings are possible for NxN board with Q colours (including that neighbours, defined as having common edge don't get same color). This ought to be pretty simple formula.
Then figure out how many of these solutions have 0 in (i,j), this should be 1/Q's fraction.
Then figure out how many of remaining solutions have 0 in (k,l) depending on manhattan distance |i-k|+|j-l|, and possibly distance to the board edge and "parity" of these distances, as in distance divisible by 2, divisible by 3, divisible by Q.
The last part is the hardest, though I think it might still be doable if you are really good at math.

Categories