I have 9 bins named A through I containing the following number of objects:
A(8), B(7), C(6), D(7), E(5), F(6), G(6), H(6), I(6)
Objects from each bin fulfill a specific role and cannot be interchanged. I am selecting one object from each bin at random forming a "team" of 9 "players":
T_ijklmnopq = {a_i, b_j, c_k, d_l, e_m, f_n, g_o, h_p, i_q}
There are 15,240,960 such combinations - a huge number. I have means of evaluating performance of each "team" via a costly objective function, F(T_ijklmnopq). Thus, I can feasibly sample a limited number of random combinations, say no more than 500 samples.
Having results of such sampling, I want to predict the most likely best combination of "players". How to do it?
Keep in mind this is different from classical team selection because there is no meaningful evaluation of F() based on individual performance. For example, "player" a_6 may be good individually, but he may not "like" e_2 and therefore the performance of "team" containing the two suffers. Conversely, three mediocre players b_1, f_5, i_2 may be a part of an awesome "team". What's know is the whole "team" performance, that's all.
One more detail: contributions of the individual roles A through I are not weighted equally. Position of, say, E may be more important than, say, H. Unfortunately, these weights are not known upfront.
The described problem must be know to combinatorial analysts, but I haven't found anything exactly like it. Linear programming solutions with known individual "player" scores do not apply here. I will be most grateful for a specific name under which this problem is known to experts.
So far I have collected 400 samples. Here is a graph of the sorted F(T) values vs. a (arbitrary) sample number to illustrate that F(T) is "reasonable".
F(T) graph of sorted samples
Related
I am currently completing an exercise book on machine learning to wet my feet so to speak in the discipline. Right now I am working on a real estate data set: each instance is a district of california and has several attributes, including the district's median income, which has been scaled and capped at 15. The median income histogram reveals that most median income values are clustered around 2 to 5, but some values go far beyond 6. The author wants to use stratified sampling, basing the strata on the median income value. He offers the next piece of code to create an income category attribute.
housing["income_cat"] = np.ceil(housing["median_income"] / 1.5)
housing["income_cat"].where(housing["income_cat"] < 5, 5.0, inplace=True)
He explains that he divides the median_income by 1.5 to limit the number of categories and that he then keeps only those categories lower than 5 and merges all other categories into category 5.
What I don't understand is
Why is it mathematically sound to divide the median_income of each instance to create the strata? What exactly does the result of this division mean? Are there other ways to calculate/limit the number of strata?
How does the division restrict the number of categories and why did he choose 1.5 as the divisor instead of a different value? How did he know which value to pick?
Why does he only want 5 categories and how did he know beforehand that there would be at least 5 categories?
Any help understanding these decisions would be greatly appreciated.
I'm also not sure if this is the StackOverFlow category I should post this question in, so if I made a mistake by doing so please let me know what might be the appropriate forum.
Thank you!
You may be the right person to analyze more on this based on your data set. But I can help you understanding stratified sampling, so that you will have an idea.
STRATIFIED SAMPLING: suppose you have a data set with consumers who eat different fruits. One feature is 'fruit type' and this feature has 10 different categories(apple,orange,grapes..etc) now if you just sample the data from data set, there is a possibility that sample data might not cover all the categories. Which is very bad when train the data. To avoid such scenario, we have a method called stratified sampling, in this probability of sampling each different category is same so that we will not miss any useful data.
Please let me know if you still have any questions, I would be very happy to help you.
I am creating a machine learning model that essentially returns the correctness of one text to another.
For example; “the cat and a dog”, “a dog and the cat”. The model needs to be able to identify that some words (“cat”/“dog”) are more important/significant than others (“a”/“the”). I am not interested in conjunction words etc. I would like to be able to tell the model which words are the most “significant” and have it determine how correct text 1 is to text 2, with the “significant” words bearing more weight than others.
It also needs to be able to recognise that phrases don’t necessarily have to be in the same order. The two above sentences should be an extremely high match.
What is the basic algorithm I should use to go about this? Is there an alternative to just creating a dataset with thousands of example texts and a score of correctness?
I am only after a broad overview/flowchart/process/algorithm.
I think TF-IDF might be a good fit to your problem, because:
Emphasis on words occurring in many documents (say, 90% of your sentences/documents contain the conjuction word 'and') is much smaller, essentially giving more weight to the more document specific phrasing (this is the IDF part).
Ordering in Term Frequency (TF) does not matter, as opposed to methods using sliding windows etc.
It is very lightweight when compared to representation oriented methods like the one mentioned above.
Big drawback: Your data, depending on the size of corpus, may have too many dimensions (the same number of dimensions as unique words), you could use stemming/lemmatization in order to mitigate this problem to some degree.
You may calculate similiarity between two TF-IDF vector using cosine similiarity for example.
EDIT: Woops, this question is 8 months old, sorry for the bump, maybe it will be of use to someone else though.
I am optimizing algorithmic strategies. In the process of choosing from a pool of many optimized strategies, I am in the phase of searching (evaluating) for robustness of the strategy.
Following the guidelines of Dr. Pardo's book "The Evaluation of Trading Strategies" in page 231 Dr. Pardo recomends, in the Numeral 3 to apply the following ratio to the optimized data:
" 3. The ratio of the total profit of all profitable simulations divided by the
total profit of all simulationsis significantly positive"
The Question: from the optimization results, I am not being able to properly understand what does Mr. Pardo means by stating "...all simulationsis significantly positive"; what does Mr. Pardo means by 'significantly positive?
a.) with 95% confidence level?
b.) with a certain p value?
c.) the relation of the average net profit of each simulation minus it' standard deviation
Even though the sentence might seem 'simple' I would REALLY like to understand what Mr. Pardo means by the statement and HOW to calculate it, in order to filter the most robust algorithmic strategies.
The aim of analyzing the optimization profile of an algorithmic simulation is to be able to filter robust strategies.
Therefore the ratio should help us to uncover if the simulation results are on the right track or not.
So, we would like to impose some 'penalties' to our results, so we can select the robust cases from those of doubtful (not robust) result.
I came to the following penalizing measures (found in the book of Mr. Pardo and other sources).
a.) we can use a market return (yearly value) as a benchmark, so all the simulations whose result are below such level, can be excluded from our analysis,
b.) some other benchmark to divide those 'robust' results from those more 'doubtful' (for example, deducing to each result one standard deviation)
From (a) and (b), we can create the ratio:
the total sum of all profitable simulations divided by the profitable results considered robust
The ratio should be greater or equal than 1.
If the ratio is equal to 1 then it means that our simulation result has given interesting results (we are analyzing the positive values in this ratio, but profitable results should always be compared to the negative results also).
If the ratio is greater from 1, then we have not reach the possible scenario, and the result should be compared with the other tests for optimizations.
While simulating trading algorithms, no result is absolute but partial and it's value is taken in relationship to what we expect from the algorithm.
If someone has a better explanation or idea or concept you might find interesting please share, I would gladly read it.
Best regards to all.
Remark on the subject
With all due respect to the subject ( published in 2008 ) the term robustness has its own meaning if-and-only-if the statement also clarifies in which particular respect is the robustness measured and against what phenomena is it to be exposed & tested the Model-under-review's response ( against what perturbances -- type and scale -- shall the Model-under-test hold its robust behaviour, measures of which were both defined and quantified a-priori the test ).
In any case, where such context of the robustness is not defined, the material, be it printed by any bold name, sounds -- and forgive me to speak in plain English -- just like a PR-story, an over-hyped e-zine headline or like a paid advertorial.
Serious quantitative model evaluations, the more if one strives to perform an optimisation ( with respect to some defined quantitative goal ), requires a more thorough insight into the subject than to axiomatically post a trivial "must-have" imperative of
large-average && small-HiLo-range && small StDev.
Any serious Quant-Modelling effort, if it were not to just spoil the consumed hundreds-of-thousands CPU core hours of deep parametric-spaces' scans, shall incorporate a serious parametrisation decision in either dimension of the main TruTrading Strategy sub-spaces --
{ aSelectPOLICY, aDetectPOLICY, anActPOLICY, anAllocatePOLICY, aTerminatePOLICY }
A failure to do so, either cripples the model or leads to a blind-belief, where it is hard to guess, whether the former or the latter is a greater of the both Quant-sins.
Remark on the cited hypothesis
The book states, without any effort to proof the construction, that:
The more robust trading strategywill have an optimization profile with a: 1. Largeaverageprofit 2. Small maximum-minimumrange3. Small standarddeviation
Is it correct?
Now kindly spend a few moments and review this 4D-animated view of a Model-under-test ( visualisation of which is reduced into just four dimensions for easier visual perception ), where none of the above stands true.
<aMouseRightCLICK>.openPictureOnAnotherTab to see full HiRes picture details
Based on contemporary state-of-art adaptive money-management practice, that fails to be correct, be it due to a poor parametrisation ( thus artificially leading the model into a rather "flat-profits" sub-space of aParamSetVectorSPACE )
or due to a principal mis-concept or a poor practice ( including the lack thereof ) of the implementation of the most powerful profit-booster ever -- the very money-management model sub-space.
Item 1 becomes insignificant at all.
Item 2 works right on the contrary to the stated postulate.
Item 3 cannot yield anything but the opposite due to 1 & 2 above.
Question after BIG edition :
I need to built a ranking using genetic algorithm, I have data like this :
P(a>b)=0.9
P(b>c)=0.7
P(c>d)=0.8
P(b>d)=0.3
now, lets interpret a,b,c,d as names of football teams, and P(x>y) is probability that x wins with y. We want to build ranking of teams, we lack some observations P(a>d),P(a>c) are missing due to lack of matches between a vs d and a vs c.
Goal is to find ordering of team names, which the best describes current situation in that four team league.
If we have only 4 teams than solution is straightforward, first we compute probabilities for all 4!=24 orderings of four teams, while ignoring missing values we have :
P(abcd)=P(a>b)P(b>c)P(c>d)P(b>d)
P(abdc)=P(a>b)P(b>c)(1-P(c>d))P(b>d)
...
P(dcba)=(1-P(a>b))(1-P(b>c))(1-P(c>d))(1-P(b>d))
and we choose the ranking with highest probability. I don't want to use any other fitness function.
My question :
As numbers of permutations of n elements is n! calculation of probabilities for all
orderings is impossible for large n (my n is about 40). I want to use genetic algorithm for that problem.
Mutation operator is simple switching of places of two (or more) elements of ranking.
But how to make crossover of two orderings ?
Could P(abcd) be interpreted as cost function of path 'abcd' in assymetric TSP problem but cost of travelling from x to y is different than cost of travelling from y to x, P(x>y)=1-P(y<x) ? There are so many crossover operators for TSP problem, but I think I have to design my own crossover operator, because my problem is slightly different from TSP. Do you have any ideas for solution or frame for conceptual analysis ?
The easiest way, on conceptual and implementation level, is to use crossover operator which make exchange of suborderings between two solutions :
CrossOver(ABcD,AcDB) = AcBD
for random subset of elements (in this case 'a,b,d' in capital letters) we copy and paste first subordering - sequence of elements 'a,b,d' to second ordering.
Edition : asymetric TSP could be turned into symmetric TSP, but with forbidden suborderings, which make GA approach unsuitable.
It's definitely an interesting problem, and it seems most of the answers and comments have focused on the semantic aspects of the problem (i.e., the meaning of the fitness function, etc.).
I'll chip in some information about the syntactic elements -- how do you do crossover and/or mutation in ways that make sense. Obviously, as you noted with the parallel to the TSP, you have a permutation problem. So if you want to use a GA, the natural representation of candidate solutions is simply an ordered list of your points, careful to avoid repitition -- that is, a permutation.
TSP is one such permutation problem, and there are a number of crossover operators (e.g., Edge Assembly Crossover) that you can take from TSP algorithms and use directly. However, I think you'll have problems with that approach. Basically, the problem is this: in TSP, the important quality of solutions is adjacency. That is, abcd has the same fitness as cdab, because it's the same tour, just starting and ending at a different city. In your example, absolute position is much more important that this notion of relative position. abcd means in a sense that a is the best point -- it's important that it came first in the list.
The key thing you have to do to get an effective crossover operator is to account for what the properties are in the parents that make them good, and try to extract and combine exactly those properties. Nick Radcliffe called this "respectful recombination" (note that paper is quite old, and the theory is now understood a bit differently, but the principle is sound). Taking a TSP-designed operator and applying it to your problem will end up producing offspring that try to conserve irrelevant information from the parents.
You ideally need an operator that attempts to preserve absolute position in the string. The best one I know of offhand is known as Cycle Crossover (CX). I'm missing a good reference off the top of my head, but I can point you to some code where I implemented it as part of my graduate work. The basic idea of CX is fairly complicated to describe, and much easier to see in action. Take the following two points:
abcdefgh
cfhgedba
Pick a starting point in parent 1 at random. For simplicity, I'll just start at position 0 with the "a".
Now drop straight down into parent 2, and observe the value there (in this case, "c").
Now search for "c" in parent 1. We find it at position 2.
Now drop straight down again, and observe the "h" in parent 2, position 2.
Again, search for this "h" in parent 1, found at position 7.
Drop straight down and observe the "a" in parent 2.
At this point note that if we search for "a" in parent one, we reach a position where we've already been. Continuing past that will just cycle. In fact, we call the sequence of positions we visited (0, 2, 7) a "cycle". Note that we can simply exchange the values at these positions between the parents as a group and both parents will retain the permutation property, because we have the same three values at each position in the cycle for both parents, just in different orders.
Make the swap of the positions included in the cycle.
Note that this is only one cycle. You then repeat this process starting from a new (unvisited) position each time until all positions have been included in a cycle. After the one iteration described in the above steps, you get the following strings (where an "X" denotes a position in the cycle where the values were swapped between the parents.
cbhdefga
afcgedbh
X X X
Just keep finding and swapping cycles until you're done.
The code I linked from my github account is going to be tightly bound to my own metaheuristics framework, but I think it's a reasonably easy task to pull the basic algorithm out from the code and adapt it for your own system.
Note that you can potentially gain quite a lot from doing something more customized to your particular domain. I think something like CX will make a better black box algorithm than something based on a TSP operator, but black boxes are usually a last resort. Other people's suggestions might lead you to a better overall algorithm.
I've worked on a somewhat similar ranking problem and followed a technique similar to what I describe below. Does this work for you:
Assume the unknown value of an object diverges from your estimate via some distribution, say, the normal distribution. Interpret your ranking statements such as a > b, 0.9 as the statement "The value a lies at the 90% percentile of the distribution centered on b".
For every statement:
def realArrival = calculate a's location on a distribution centered on b
def arrivalGap = | realArrival - expectedArrival |
def fitness = Σ arrivalGap
Fitness function is MIN(fitness)
FWIW, my problem was actually a bin-packing problem, where the equivalent of your "rank" statements were user-provided rankings (1, 2, 3, etc.). So not quite TSP, but NP-Hard. OTOH, bin-packing has a pseudo-polynomial solution proportional to accepted error, which is what I eventually used. I'm not quite sure that would work with your probabilistic ranking statements.
What an interesting problem! If I understand it, what you're really asking is:
"Given a weighted, directed graph, with each edge-weight in the graph representing the probability that the arc is drawn in the correct direction, return the complete sequence of nodes with maximum probability of being a topological sort of the graph."
So if your graph has N edges, there are 2^N graphs of varying likelihood, with some orderings appearing in more than one graph.
I don't know if this will help (very brief Google searches did not enlighten me, but maybe you'll have more success with more perseverance) but my thoughts are that looking for "topological sort" in conjunction with any of "probabilistic", "random", "noise," or "error" (because the edge weights can be considered as a reliability factor) might be helpful.
I strongly question your assertion, in your example, that P(a>c) is not needed, though. You know your application space best, but it seems to me that specifying P(a>c) = 0.99 will give a different fitness for f(abc) than specifying P(a>c) = 0.01.
You might want to throw in "Bayesian" as well, since you might be able to start to infer values for (in your example) P(a>c) given your conditions and hypothetical solutions. The problem is, "topological sort" and "bayesian" is going to give you a whole bunch of hits related to markov chains and markov decision problems, which may or may not be helpful.
I have read through various papers on the 'Balls and Bins' problem and it seems that if a hash function is working right (ie. it is effectively a random distribution) then the following should/must be true if I hash n values into a hash table with n slots (or bins):
Probability that a bin is empty, for large n is 1/e.
Expected number of empty bins is n/e.
Probability that a bin has k balls is <= 1/ek! (corrected).
Probability that a bin has at least k collisions is <= ((e/k)**k)/e (corrected).
These look easy to check. But the max-load test (the maximum number of collisions with high probability) is usually stated vaguely.
Most texts state that the maximum number of collisions in any bin is O( ln(n) / ln(ln(n)) ).
Some say it is 3*ln(n) / ln(ln(n)). Other papers mix ln and log - usually without defining them, or state that log is log base e and then use ln elsewhere.
Is ln the log to base e or 2 and is this max-load formula right and how big should n be to run a test?
This lecture seems to cover it best, but I am no mathematician.
http://pages.cs.wisc.edu/~shuchi/courses/787-F07/scribe-notes/lecture07.pdf
BTW, with high probability seems to mean 1 - 1/n.
That is a fascinating paper/lecture-- makes me wish I had taken some formal algorithms class.
I'm going to take a stab at some answers here, based on what I've just read from that, and feel free to vote me down. I'd appreciate a correction, though, rather than just a downvote :) I'm also going to use n and N interchangeably here, which is a big no-no in some circles, but since I'm just copy-pasting your formulae, I hope you'll forgive me.
First, the base of the logs. These numbers are given as big-O notation, not as absolute formulae. That means that you're looking for something 'on the order of ln(n) / ln(ln(n))', not with an expectation of an absolute answer, but more that as n gets bigger, the relationship of n to the maximum number of collisions should follow that formula. The details of the actual curve you can graph will vary by implementation (and I don't know enough about the practical implementations to tell you what's a 'good' curve, except that it should follow that big-O relationship). Those two formulae that you posted are actually equivalent in big-O notation. The 3 in the second formula is just a constant, and is related to a particular implementation. A less efficient implementation would have a bigger constant.
With that in mind, I would run empirical tests, because I'm a biologist at heart and I was trained to avoid hard-and-fast proofs as indications of how the world actually works. Start with N as some number, say 100, and find the bin with the largest number of collisions in it. That's your max-load for that run. Now, your examples should be as close as possible to what you expect actual users to use, so maybe you want to randomly pull words from a dictionary or something similar as your input.
Run that test many times, at least 30 or 40. Since you're using random numbers, you'll need to satisfy yourself that the average max-load you're getting is close to the theoretical 'expectation' of your algorithm. Expectation is just the average, but you'll still need to find it, and the tighter your std dev/std err about that average, the more you can say that your empirical average matches the theoretical expectation. One run is not enough, because a second run will (most likely) give a different answer.
Then, increase N, to say, 1000, 10000, etc. Increase it logarithmically, because your formula is logarithmic. As your N increases, your max-load should increase on the order of ln(n) / ln(ln(n)). If it increases at a rate of 3*ln(n) / ln(ln(n)), that means that you're following the theory that they put forth in that lecture.
This kind of empirical test will also show you where your approach breaks down. It may be that your algorithm works well for N < 10 million (or some other number), but above that, it starts to collapse. Why could that be? Maybe you have some limitation to 32 bits in your code without realizing it (ie, using a 'float' instead of a 'double'), or some other implementation detail. These kinds of details let you know where your code will work well in practice, and then as your practical needs change, you can modify your algorithm. Maybe making the algorithm work for very large datasets makes it very inefficient for very small ones, or vice versa, so pinpointing that tradeoff will help you further characterize how you could adapt your algorithm to particular situations. Always a useful skill to have.
EDIT: a proof of why the base of the log function doesn't matter with big-O notation:
log N = log_10 (N) = log_b (N)/log_b (10)= (1/log_b(10)) * log_b(N)
1/log_b(10) is a constant, and in big-O notation, constants are ignored. Base changes are free, which is why you're encountering such variation in the papers.
Here is a rough start to the solution of this problem involving uniform distributions and maximum load.
Instead of bins and balls or urns or boxes or buckets or m and n, people (p) and doors (d) will be used as designations.
There is an exact expected value for each of the doors given a certain number of people. For example, with 5 people and 5 doors, the expected maximum door is exactly 1.2864 {(1429-625) / 625} above the mean (p/d) and the minimum door is exactly -0.9616 {(24-625) / 625} below the mean. The absolute value of the highest door's distance from the mean is a little larger than the smallest door's because all of the people could go through one door, but no less than zero can go through one of the doors. With large numbers of people (p/d > 3000), the difference between the absolute value of the highest door's distance from the mean and the lowest door's becomes negligible.
For an odd number of doors, the center door is essentially zero and is not scalable, but all of the other doors are scalable from certain values representing p=d. These rounded values for d=5 are:
-1.163 -0.495 0* 0.495 1.163
* slowly approaching zero from -0.12
From these values, you can compute the expected number of people for any count of people going through each of the 5 doors, including the maximum door. Except for the middle ordered door, the difference from the mean is scalable by sqrt(p/d).
So, for p=50,000 and d=5:
Expected number of people going through the maximum door, which could be any of the 5 doors, = 1.163 * sqrt(p/d) + p/d.
= 1.163 * sqrt(10,000) + 10,000 = 10,116.3
For p/d < 3,000, the result from this equation must be slightly increased.
With more people, the middle door slowly becomes closer and closer to zero from -0.11968 at p=100 and d=5. It can always be rounded up to zero and like the other 4 doors has quite a variance.
The values for 6 doors are:
-1.272 -0.643 -0.202 0.202 0.643 1.272
For 1000 doors, the approximate values are:
-3.25, -2.95, -2.79 … 2.79, 2.95, 3.25
For any d and p, there is an exact expected value for each of the ordered doors. Hopefully, a good approximation (with a relative error < 1%) exists. Some professor or mathematician somewhere must know.
For testing uniform distribution, you will need a number of averaged ordered sessions (750-1000 works well) rather than a greater number of people. No matter what, the variances between valid sessions are great. That's the nature of randomness. Collisions are unavoidable. *
The expected values for 5 and 6 doors were obtained by sheer brute force computation using 640 bit integers and averaging the convergence of the absolute values of corresponding opposite doors.
For d=5 and p=170:
-6.63901 -2.95905 -0.119342 2.81054 6.90686
(27.36099 31.04095 33.880658 36.81054 40.90686)
For d=6 and p=108:
-5.19024 -2.7711 -0.973979 0.734434 2.66716 5.53372
(12.80976 15.2289 17.026021 18.734434 20.66716 23.53372)
I hope that you may evenly distribute your data.
It's almost guaranteed that all of George Foreman's sons or some similar situation will fight against your hash function. And proper contingent planning is the work of all good programmers.
After some more research and trial-and-error I think I can provide something part way to to an answer.
To start off, ln and log seem to refer to log base-e if you look into the maths behind the theory. But as mmr indicated, for the O(...) estimates, it doesn't matter.
max-load can be defined for any probability you like. The typical formula used is
1-1/n**c
Most papers on the topic use
1-1/n
An example might be easiest.
Say you have a hash table of 1000 slots and you want to hash 1000 things. Say you also want to know the max-load with a probability of 1-1/1000 or 0.999.
The max-load is the maximum number of hash values that end up being the same - ie. collisions (assuming that your hash function is good).
Using the formula for the probability of getting exactly k identical hash values
Pr[ exactly k ] = ((e/k)**k)/e
then by accumulating the probability of exactly 0..k items until the total equals or exceeds 0.999 tells you that k is the max-load.
eg.
Pr[0] = 0.37
Pr[1] = 0.37
Pr[2] = 0.18
Pr[3] = 0.061
Pr[4] = 0.015
Pr[5] = 0.003 // here, the cumulative total is 0.999
Pr[6] = 0.0005
Pr[7] = 0.00007
So, in this case, the max-load is 5.
So if my hash function is working well on my set of data then I should expect the maxmium number of identical hash values (or collisions) to be 5.
If it isn't then this could be due to the following reasons:
Your data has small values (like short strings) that hash to the same value. Any hash of a single ASCII character will pick 1 of 128 hash values (there are ways around this. For example you could use multiple hash functions, but slows down hashing and I don't know much about this).
Your hash function doesn't work well with your data - try it with random data.
Your hash function doesn't work well.
The other tests I mentioned in my question also are helpful to see that your hash function is running as expected.
Incidentally, my hash function worked nicely - except on short (1..4 character) strings.
I also implemented a simple split-table version which places the hash value into the least used slot from a choice of 2 locations. This more than halves the number of collisions and means that adding and searching the hash table is a little slower.
I hope this helps.