I've been thinking a lot lately when driving my car - inside the ECU there is a memory module with pre-calculated values for almost anything. For example, the ECU can calculate how much fuel to inject based on several readings such as throttle position, current RPM's, etc. When people remap their cars they change the predefined values which in turn changes the output calculated in realtime by the ECU. Let's keep it simple and imagine we have 2 parameters we constantly juggle around on a predefined 2D graph. We have 4 reference points: A1(2000 RPM - 200 foo units), A2(3000 RPM - 270 foo units), A3(4000 RPM - 350 foo units), A4(5000 RPM - 400 foo units). So the question I'm struggling with is how can you calculate the exact amount of foo units on let's say 3650 RPM in realtime on "slow" hardware without any errors or delays. I'd love to see some C style pseudo code on how it could be implemented logic-wise to run efficiently. The first thing that comes to my mind are 2 arrays (a matrix), but things get messy when you account for multiple variables making a difference on the final outcome. I'd like to experiment with this and try to write a small program to do this kind of math, but I'm stuck on choosing the clean, sane way of representing and manipulating values...
Sorry for no formatting, wrote this post on my phone!
Related
and thanks for reading my thread.
I have read some of the previous posts on formatting/normalising input data for a Neural Network, but cannot find something that addresses my queries specifically. I apologise for the long post.
I am attempting to build a radial basis function network for analysing horse racing data. I realise that this has been done before, but the data that I have is "special" and I have a keen interest in racing/sportsbetting/programming so would like to give it a shot!
Whilst I think I understand the principles for the RBFN itself, I am having some trouble understanding the normalisation/formatting/scaling of the input data so that it is presented in a "sensible manner" for the network, and I am not sure how I should formulate the output target values.
For example, in my data I look at the "Class change", which compares the class of race that the horse is running in now compared to the race before, and can have a value between -5 and +5. I expect that I need to rescale these to between -1 and +1 (right?!), but I have noticed that many more runners have a class change of 1, 0 or -1 than any other value, so I am worried about "over-representation". It is not possible to gather more data for the higher/lower class changes because thats just 'the way the data comes'. Would it be best to use the data as-is after scaling, or should I trim extreme values, or something else?
Similarly, there are "continuous" inputs - like the "Days Since Last Run". It can have a value between 1 and about 1000, but values in the range of 10-40 vastly dominate. I was going to scale these values to be between 0 and 1, but even if I trim the most extreme values before scaling, I am still going to have a huge representation of a certain range - is this going to cause me an issue? How are problems like this usually dealt with?
Finally, I am having trouble understanding how to present the "target" values for training to the network. My existing results data has the "win/lose" (0 or 1?) and the odds at which the runner won or lost. If I just use the "win/lose", it treats all wins and loses the same when really they're not - I would be quite happy with a network that ignored all the small winners but was highly profitable from picking 10-1 shots. Similarly, a network could be forgiven for "losing" on a 20-1 shot but losing a bet at 2/5 would be a bad loss. I considered making the results (+1 * odds) for a winner and (-1 / odds) for a loser to capture the issue above, but this will mean that my results are not a continuous function as there will be a "discontinuity" between short price winners and short price losers.
Should I have two outputs to cover this - one for bet/no bet, and another for "stake"?
I am sorry for the flood of questions and the long post, but this would really help me set off on the right track.
Thank you for any help anyone can offer me!
Kind regards,
Paul
The documentation that came with your RBFN is a good starting point to answer some of these questions.
Trimming data aka "clamping" or "winsorizing" is something I use for similar data. For example "days since last run" for a horse could be anything from just one day to several years but tends to centre in the region of 20 to 30 days. Some experts use a figure of say 63 days to indicate a "spell" so you could have an indicator variable like "> 63 =1 else 0" for example. One clue is to look at outliers say the upper or lower 5% of any variable and clamp these.
If you use odds/dividends anywhere make sure you use the probabilities ie 1/(odds+1) and a useful idea is to normalize these to 100%.
The odds or parimutual prices tend to swamp other predictors so one technique is to develop separate models, one for the market variables (the market model) and another for the non-market variables (often called the "fundamental" model).
Im trying to analyse data from cycle accidents in the UK to find statistical black spots. Here is the example of the data from another website. http://www.cycleinjury.co.uk/map
I am currently using SQLite to ~100k store lat / lon locations. I want to group nearby locations together. This task is called cluster analysis.
I would like simplify the dataset by ignoring isolated incidents and instead only showing the origin of clusters where more than one accident have taken place in a small area.
There are 3 problems I need to overcome.
Performance - How do I ensure finding nearby points is quick. Should I use SQLite's implementation of an R-Tree for example?
Chains - How do I avoid picking up chains of nearby points?
Density - How to take cycle population density into account? There is a far greater population density of cyclist in london then say Bristol, therefore there appears to be a greater number of backstops in London.
I would like to avoid 'chain' scenarios like this:
Instead I would like to find clusters:
London screenshot (I hand drew some clusters)...
Bristol screenshot - Much lower density - the same program ran over this area might not find any blackspots if relative density was not taken into account.
Any pointers would be great!
Well, your problem description reads exactly like the DBSCAN clustering algorithm (Wikipedia). It avoids chain effects in the sense that it requires them to be at least minPts objects.
As for the differences in densities across, that is what OPTICS (Wikipedia) is supposed do solve. You may need to use a different way of extracting clusters though.
Well, ok, maybe not 100% - you maybe want to have single hotspots, not areas that are "density connected". When thinking of an OPTICS plot, I figure you are only interested in small but deep valleys, not in large valleys. You could probably use the OPTICS plot an scan for local minima of "at least 10 accidents".
Update: Thanks for the pointer to the data set. It's really interesting. So I did not filter it down to cyclists, but right now I'm using all 1.2 million records with coordinates. I've fed them into ELKI for analysis, because it's really fast, and it actually can use the geodetic distance (i.e. on latitude and longitude) instead of Euclidean distance, to avoid bias. I've enabled the R*-tree index with STR bulk loading, because that is supposed to help to get the runtime down a lot. I'm running OPTICS with Xi=.1, epsilon=1 (km) and minPts=100 (looking for large clusters only). Runtime was around 11 Minutes, not too bad. The OPTICS plot of course would be 1.2 million pixels wide, so it's not really good for full visualization anymore. Given the huge threshold, it identified 18 clusters with 100-200 instances each. I'll try to visualize these clusters next. But definitely try a lower minPts for your experiments.
So here are the major clusters found:
51.690713 -0.045545 a crossing on A10 north of London just past M25
51.477804 -0.404462 "Waggoners Roundabout"
51.690713 -0.045545 "Halton Cross Roundabout" or the crossing south of it
51.436707 -0.499702 Fork of A30 and A308 Staines By-Pass
53.556186 -2.489059 M61 exit to A58, North-West of Manchester
55.170139 -1.532917 A189, North Seaton Roundabout
55.067229 -1.577334 A189 and A19, just south of this, a four lane roundabout.
51.570594 -0.096159 Manour House, Picadilly Line
53.477601 -1.152863 M18 and A1(M)
53.091369 -0.789684 A1, A17 and A46, a complex construct with roundabouts on both sides of A1.
52.949281 -0.97896 A52 and A46
50.659544 -1.15251 Isle of Wight, Sandown.
...
Note, these are just random points taken from the clusters. It may be sensible to compute e.g. cluster center and radius instead, but I didn't do that. I just wanted to get a glimpse of that data set, and it looks interesting.
Here are some screenshots, with minPts=50, epsilon=0.1, xi=0.02:
Notice that with OPTICS, clusters can be hierarchical. Here is a detail:
First, your example is quite misleading. You have two different sets of data, and you don't control the data. If it appears in a chain, then you will get a chain out.
This problem is not exactly suitable for a database. You'll have to write code or find a package that implements this algorithm on your platform.
There are many different clustering algorithms. One, k-means, is an iterative algorithm where you look for a fixed number of clusters. k-means requires a few complete scans of the data, and voila, you have your clusters. Indexes are not particularly helpful.
Another, which is usually appropriate on slightly smaller data sets, is hierarchical clustering -- you put the two closest things together, and then build the clusters. An index might be helpful here.
I recommend though that you peruse a site such as kdnuggets in order to see what software -- free and otherwise -- is available.
Y
I have 6 parameters for which I know maxi and mini values. I have a complex function that includes the 6 parameters and return a 7th value (say Y). I say complex because Y is not directly related to the 6 parameters; there are many embeded functions in between.
I would like to find the combination of the 6 parameters which returns the highest Y value. I first tried to calculate Y for every combination by constructing an hypercube but I have not enough memory in my computer. So I am looking for kinds of markov chains which progress in the delimited parameter space, and are able to overpass local peaks.
when I give one combination of the 6 parameters, I would like to know the highest local Y value. I tried to write a code with an iterative chain like a markov's one, but I am not sure how to process when the chain reach an edge of the parameter space. Obviously, some algorythms should already exist for this.
Question: Does anybody know what are the best functions in R to do these two things? I read that optim() could be appropriate to find the global peak but I am not sure that it can deal with complex functions (I prefer asking before engaging in a long (for me) process of code writing). And fot he local peaks? optim() should not be able to do this
In advance, thank you for any lead
Julien from France
Take a look at the Optimization and Mathematical Programming Task View on CRAN. I've personally found the differential evolution algorithm to be very fast and robust. It's implemented in the DEoptim package. The rgenoud package is another good candidate.
I like to use the Metropolis-Hastings algorithm. Since you are limiting each parameter to a range, the simple thing to do is let your proposal distribution simply be uniform over the range. That way, you won't run off the edges. It won't be fast, but if you let it run long enough, it will do a good job of sampling your space. The samples will congregate at each peak, and will spread out around them in a way that reflects the local curvature.
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.
Recently I was faced with this interview question (K-Means Clustering solution). The design I came up with did not meet the expectations of the interviewer (to put simply I didnt get the job because I lost to another candidate on this design problem). I am wondering how many different / efficient / simply solutions can the SO community come up with (by doing this I am hoping to hone my skills):
To implement a simple algorithm to cluster people according to their weight and height. The
data set includes a list of people with their weights and heights like so:
Person Weight Height
(kg) (inches)
Person 1 70 70
Person 2 75 80
Person 3 120 85
You can plot the data as a 2 dimensional data. Weight being one dimension and height being
the other dimension. Weight can range from a minimum of 50kg to 150kg. Height can range
from a minimum of 38inches to 90inches
Algorithm:
The algorithm (called K-means clustering) will cluster data into K groups goes as such:
Start with K clusters. Each cluster is defined by its center point which will start of as
random weight and random height. Pick random numbers from within the
corresponding ranges defined above.
For each person
Calculate distance to center of each cluster using formula
distance = sqrt(pow((wperson−wcenter), 2) + (pow(hperson−hcenter),2))
where wperson = weight of person,
hperson = height of person
wcenter = weight of cluster center point,
hcenter = height of cluster center point
Assign the person to the cluster with the shortest distance to center point of cluster
After end of step 2, you will end up with K clusters each assigned with a set of people
For each cluster, set the weight and height of the center point to the average of the
people in the cluster
wcenter = (sum of weight of each person in cluster)/(number of people in cluster)
hcenter = (sum of height of each person in cluster)/number of people in cluster)
Repeat steps 2 to 5 for 1000 iterations, then print out following information for each
cluster.
weight and height of center of cluster.
list of people in cluster.
I am not looking for a implementation/solution but for a high level design. can you list the interfaces / classes etc.
I dont want to give my solution now, but will post it later in the day?
This is my attempt at the design. I only show the static diagram since the algorithm is pretty much laid out already. I would have a plan to have a visitor for the representation of the clusters, could allow different types of output (xml, strings, csv..etc). Maybe the visitor is overkill, if it was then I'd just have something like a ToString method that could be overridden.
Note: the Cluster creates a CenterClusterItem on the SetCenter and FindNewCenter methods. The CenterClusterItem is not a PersonClusterItem, it just holds the same amount of AClusterValues as a PersonClusterItem would (since the average isn't really a person).
Also, I forgot to make a method on the KCluster to begin the process, but that's implied.
Class Diagram http://img11.imageshack.us/img11/499/kcluster.png
Well, I would first tackle all the constants/magic numbers that reduce the reusability of the algorithm:
instead of a fixed number of iterations, use a stopping criterion (e.g., if clusters don't change too much, terminate)
don't restrict yourself to 2-dim data, use vectors
let the user define the number of clusters to be found
Then, you could hide some specifics behind interfaces, e.g. the distance might be calculated differently (for example, it might at some point have to cope with values other than double).
On the other hand, if you really have this simple problem, some of these generalizations might well be overkill - but that's what I would discuss with someone telling me to implement this algorithm.
You can create the following classes:
Person to store data about persons and centers. Properties: id, weight and height. Method: calculateDistance
Cluster to store one center and a list of persons: Properties: center and list of Person. Method: calculateCenter.
KCluster to hold your algorithm and store a list of clusters: Property: list of Cluster. Methods: generateClusters.
I'm not sure what your question actually is, the steps you point out effectively define the algorithm you're talking about.
A better idea may be to include exactly what you did then people can give you some hints / tips as to where you might have gone wrong or what they would have done differently.
That sounds like a really good way to do it. K-means will usually converge quickly (though not necessarily to the global optimum), so my one suggestion would be to run the algorithm until no more changes occur, rather than a fixed number of 1000 iterations. You could then repeat the entire process a few times with different random starting points.
One weakness of k-means is that it does require you to specify (i.e. guess) an appropriate value for k up-front. I think you would get points for asking the interviewer what an appropriate value for k would be, or, if there is no way to know, describing some goodness-of-fit measure and then calculating that measure for different values of k to find a "just low enough" value.