I am looking for a method to find the best parameters for a simulation. It's about break-shots in billiards / pool. A shot is defined by 7 parameters, I can simulate the shot and then rate the outcome and I would like to compute the best parameters.
I have found the following link here:
Multiple parameter optimization with lots of local minima
suggesting 4 kinds of algorithms. In the pool simulator I am using, the shots are altered by a little random value each time it is simulated. If I simulate the same shot twice, the outcome will be different. So I am looking for an algorithm like the ones in the link above, only with the addition of a stochastical element, optimizing for the 7 parameters that will on average yield the best parameters, i.e. a break shot that most likely will be a success. My initial idea was simulating the shot 100 or 1000 times and just take the average as rating for the algorithms above, but I still feel like there is a better way. Does anyone have an idea?
The 7 parameters are continuous but within different ranges (one from 0 to 10, another from 0.0 to 0.028575 and so on).
Thank you
At least for some of the algorithms, simulating the same shot repeatedly might not be neccessary. As long as your alternatives have some form of momentum, like in the swarm simulation approach, you can let that be affected by the outcome of each individual simulation. In that case, a single unlucky simulation would slow the movement in parameter space only slightly, whereas a serious loss of quality should be enough to stop and reverse the movement. Thos algorithms which don't use momentum might be tweaked to have momentum. If not, then repeated simulation seems the best approach. Unless you can get your hands on the internals of the simulator, and rate the shot as a whole without having to simulate it over and over again.
You can use the algorithms you mentioned in your non-deterministic scenario with independent stochastic runs. Your idea with repeated simulations is good, you can read more about how many repeats you might have to consider for your simulations (unfortunately, there is no trivial answer). If you are not so much into maths, and the runs go fast, do 1.000 repeats, then 10.000 repeats, and see if the results differ largely. If yes, you have to collect more samples, if not, you are probably on the safe side (the central limit theorem states that the results converge).
Further, do not just consider the average! Make sure to look into the standard deviation for each algorithm's results; you might want to use box plots to compare their quartiles. If you rely on the average only, you could pick an algorithm that produces very varying results, sometimes excellent, sometimes terrible in performance.
I don't know what language you are using, but if you use Java, I am maintaining a tool that could simplify your "monte carlo" style experiments.
Related
So I have a very simple NN script written in Tensorflow, and I am having a hard time trying to trace down where some "randomness" is coming in from.
I have recorded the
Weights,
Gradients,
Logits
of my network as I train, and for the first iteration, it is clear that everything starts off the same. I have a SEED value both for how data is read in, and a SEED value for initializing the weights of the net. Those I never change.
My problem is that on say the second iteration of every re-run I do, I start to see the gradients diverge, (by a small amount, like say, 1e-6 or so). However over time, this of course leads to non-repeatable behaviour.
What might the cause of this be? I dont know where any possible source of randomness might be coming from...
Thanks
There's a good chance you could get deterministic results if you run your network on CPU (export CUDA_VISIBLE_DEVICES=), with single-thread in Eigen thread pool (tf.Session(config=tf.ConfigProto(intra_op_parallelism_threads=1)), one Python thread (no multi-threaded queue-runners that you get from ops like tf.batch), and a single well-defined operation order. Also using inter_op_parallelism_threads=1 may help in some scenarios.
One issue is that floating point addition/multiplication is non-associative, so one fool-proof way to get deterministic results is to use integer arithmetic or quantized values.
Barring that, you could isolate which operation is non-deterministic, and try to avoid using that op. For instance, there's tf.add_n op, which doesn't say anything about the order in which it sums the values, but different orders produce different results.
Getting deterministic results is a bit of an uphill battle because determinism is in conflict with performance, and performance is usually the goal that gets more attention. An alternative to trying to have exact same numbers on reruns is to focus on numerical stability -- if your algorithm is stable, then you will get reproducible results (ie, same number of misclassifications) even though exact parameter values may be slightly different
The tensorflow reduce_sum op is specifically known to be non-deterministic. Furthermore, reduce_sum is used for calculating bias gradients.
This post discusses a workaround to avoid using reduce_sum (ie taking the dot product of any vector w/ a vector of all 1's is the same as reduce_sum)
I have faced the same problem..
The working solution for me was to:
1- use tf.set_random_seed(1) in order to make all tf functions have the same seed every new run
2- Training the model using CPU not the GPU to avoid GPU non-deterministic operations due to precision.
I would like to know the difference between uncertainty and randomness in mathematical fashion. I tried to find it but I get confused , as some people said they are the same? But can any one provide me logical reasoning behind it. If they are not same then please explain it why?
Don't get too hung up on it.
People use different words in different situations.
It's not so much that they have different meanings, as that their meanings are situation-dependent.
Randomness is just a fuzzy general term meaning something is random.
In statistics, uncertainty is used to mean that some property of a distribution, such as its mean, is itself unknown but can be given a distribution.
For example, suppose you want to know the average weight of all people.
You could find it out exactly if you could go around to all people, get their weight, add it all up, and divide by the number of people.
But that's too hard to do, so suppose you just pick 10 people at random and get their average weight, and pretend it's the same as the average of everybody.
That's called the sample mean, but you know it isn't accurate.
It has what is called a standard error, meaning it has uncertainty.
In fact, if you were to do that experiment many times over with different people, you would get a different sample mean every time, and those sample means would themselves form a bell-shaped distribution, the standard deviation of which would be called the standard error, representing its uncertainty.
In general, if you increased the number of people you look at by a factor of 100, you can reduce the standard error, the uncertainty, by a factor of 10.
I bet you can tell that people who take polls for a living care about this stuff very much.
EDIT for the downvoter: In case the downvote is because this doesn't look like a stackoverflow question or answer,
I've made a point of advocating the random pausing method of profiling.
Profiling in large part is perceived to be about measuring (statistically) the time that programming constructs are responsible for.
Often people are inhibited from using that method because they are afraid the results have too much uncertainty.
This post gets very specific about what that uncertainty actually is.
It shows that the bogey-man fear of uncertainty has the effect of preventing people from finding really substantial speedups in their code.
So naivete' about statistics is definitely a serious programming problem.
My view looks at a scenario using three different coloured balls:
I love some of the answers given here. My own view, based on my current research, is that these are two distinct terms. Uncertainty refers to not knowing in advance which ball could be selected when a person, for instance, is given a chance to select one ball from three different coloured balls.
This remains true when each ball has an equal chance of being selected i.e. equal probabilities. However, things soon get complex when each ball has it's own distinct probability. Chances are that the one with the highest probability will be selected. This seems especially true in algorithm development which would almost always select the highest probability compromising the meaning of randomness.
Having said all of this - I believe these concepts remain confusing which has just made me realise the time I need to dedicate on clearly distinguishing between the two to make sure my current research is not confusing. My own predicament is that I need to work on stochastic vs deterministic views. Based on the current view stochastic would be more uncertain than random whereas deterministic would be more probability based i.e. knowing for certain that the highest probability would be chosen; but this seems very far from the truth.
It seems as if uncertainty holds until just before a ball is selected/touched and soon looses its meaning as soon as the ball is picked which should result to its probability being revised. I personally think the terms have theoretical differences which perhaps allows them to be used interchangeably.
Uncertainty in math and science typically means there are a lack of facts, or the facts are unobtainable. Weather forecasting is a great example of uncertainty.
Randomness has many definitions. Commonly it's used in probability / statistics as a measure or quantification of uncertainty. So in my weather example, a 30% chance of rain is a measure of uncertainty. The more general definition (which also applies to math / science) is unpredictable, or lack of order.
There is definitely a fuzzy distinction between the two.
According to the Bayesian interpretation of probability, uncertainty and randomness are just two names for the same thing.
If an experiment is random, then it is uncertain to you. If something is uncertain to you, then it has the randomness property.
I have the following problem: I have 2 signals over time. They are from the same source so they should be the same. I want to check if they really are.
Complications:
they may be measured with different sample rates
the start / end time do not correlate. The measurement does not start at the same time and end at the same time.
there may be an time offset between the two signals.
My thoughts go along Fourier transformation, convolution and statistical methods for comparison. Can someone post me some links where I can find more information on how to handle this?
You can easily correct for the phase by just shifting them so their centers of mass line up. (Or alternatively, in the Fourier domain just multiplying by the inverse of the phase of the first coefficient.)
Similarly, if you want to line up the images given only partial data, you can just cross correlate and take the maximal value (which is again easy to do in the Fourier domain).
That leaves the only tricky part of this process as dealing with the sampling rates. Now if you know a-priori what the sample rates are, (and if they are related by a rational number), you can just use sinc interpolation/downsampling to rescale them to a common sampling rate:
https://ccrma.stanford.edu/~jos/st/Bandlimited_Interpolation_Time_Limited_Signals.html
If you don't know the sampling rate, you may be a bit screwed. Technically, you can try just brute forcing over all the different rescalings of your signal, but doing this tends to be either slow or else give mediocre results.
As a last suggestion, if you just want to match sounds exactly you can try using the cepstrum and verifying that the peaks of the signal are close enough to within some tolerance. This type of analysis is used a lot in sound and speech recognition, with some refinements to make it operate a bit more locally. It tends to work best with frequency modulated data like speech and music:
http://en.wikipedia.org/wiki/Cepstrum
Fourier transformation does sound like the right way.
There is too much mathematical information for me to just start explaining here so if you really wanna know what's going on with that (cause I don't think you can just use FT without understanding it) you should use this reference from MIT OpenCourseWare: http://ocw.mit.edu/courses/mathematics/18-103-fourier-analysis-theory-and-applications-spring-2004/lecture-notes/
Hope it helped.
If you are working with a linux box and the waveforms that need to be processed have already been recorded, you can try to use the file command to display details about the recording. It gives you the sampling rate when it is invoked on a wav file, though I am not sure what format you are recording in.
If the signals are time-shifted with respect to each other, you may try to convolve one with a delta function with increasing delays and then comparing. On MATLAB, conv and all should be good enough.
These are just 'crude' attempts (almost like hacking at the problem). There may be algorithms that are shift-invariant that may do a better job.
Hope that helps.
I'm looking for algorithms to find a "best" set of parameter values. The function in question has a lot of local minima and changes very quickly. To make matters even worse, testing a set of parameters is very slow - on the order of 1 minute - and I can't compute the gradient directly.
Are there any well-known algorithms for this kind of optimization?
I've had moderate success with just trying random values. I'm wondering if I can improve the performance by making the random parameter chooser have a lower chance of picking parameters close to ones that had produced bad results in the past. Is there a name for this approach so that I can search for specific advice?
More info:
Parameters are continuous
There are on the order of 5-10 parameters. Certainly not more than 10.
How many parameters are there -- eg, how many dimensions in the search space? Are they continuous or discrete - eg, real numbers, or integers, or just a few possible values?
Approaches that I've seen used for these kind of problems have a similar overall structure - take a large number of sample points, and adjust them all towards regions that have "good" answers somehow. Since you have a lot of points, their relative differences serve as a makeshift gradient.
Simulated
Annealing: The classic approach. Take a bunch of points, probabalistically move some to a neighbouring point chosen at at random depending on how much better it is.
Particle
Swarm Optimization: Take a "swarm" of particles with velocities in the search space, probabalistically randomly move a particle; if it's an improvement, let the whole swarm know.
Genetic Algorithms: This is a little different. Rather than using the neighbours information like above, you take the best results each time and "cross-breed" them hoping to get the best characteristics of each.
The wikipedia links have pseudocode for the first two; GA methods have so much variety that it's hard to list just one algorithm, but you can follow links from there. Note that there are implementations for all of the above out there that you can use or take as a starting point.
Note that all of these -- and really any approach to this large-dimensional search algorithm - are heuristics, which mean they have parameters which have to be tuned to your particular problem. Which can be tedious.
By the way, the fact that the function evaluation is so expensive can be made to work for you a bit; since all the above methods involve lots of independant function evaluations, that piece of the algorithm can be trivially parallelized with OpenMP or something similar to make use of as many cores as you have on your machine.
Your situation seems to be similar to that of the poster of Software to Tune/Calibrate Properties for Heuristic Algorithms, and I would give you the same advice I gave there: consider a Metropolis-Hastings like approach with multiple walkers and a simulated annealing of the step sizes.
The difficulty in using a Monte Carlo methods in your case is the expensive evaluation of each candidate. How expensive, compared to the time you have at hand? If you need a good answer in a few minutes this isn't going to be fast enough. If you can leave it running over night, it'll work reasonably well.
Given a complicated search space, I'd recommend a random initial distributed. You final answer may simply be the best individual result recorded during the whole run, or the mean position of the walker with the best result.
Don't be put off that I was discussing maximizing there and you want to minimize: the figure of merit can be negated or inverted.
I've tried Simulated Annealing and Particle Swarm Optimization. (As a reminder, I couldn't use gradient descent because the gradient cannot be computed).
I've also tried an algorithm that does the following:
Pick a random point and a random direction
Evaluate the function
Keep moving along the random direction for as long as the result keeps improving, speeding up on every successful iteration.
When the result stops improving, step back and instead attempt to move into an orthogonal direction by the same distance.
This "orthogonal direction" was generated by creating a random orthogonal matrix (adapted this code) with the necessary number of dimensions.
If moving in the orthogonal direction improved the result, the algorithm just continued with that direction. If none of the directions improved the result, the jump distance was halved and a new set of orthogonal directions would be attempted. Eventually the algorithm concluded it must be in a local minimum, remembered it and restarted the whole lot at a new random point.
This approach performed considerably better than Simulated Annealing and Particle Swarm: it required fewer evaluations of the (very slow) function to achieve a result of the same quality.
Of course my implementations of S.A. and P.S.O. could well be flawed - these are tricky algorithms with a lot of room for tweaking parameters. But I just thought I'd mention what ended up working best for me.
I can't really help you with finding an algorithm for your specific problem.
However in regards to the random choosing of parameters I think what you are looking for are genetic algorithms. Genetic algorithms are generally based on choosing some random input, selecting those, which are the best fit (so far) for the problem, and randomly mutating/combining them to generate a next generation for which again the best are selected.
If the function is more or less continous (that is small mutations of good inputs generally won't generate bad inputs (small being a somewhat generic)), this would work reasonably well for your problem.
There is no generalized way to answer your question. There are lots of books/papers on the subject matter, but you'll have to choose your path according to your needs, which are not clearly spoken here.
Some things to know, however - 1min/test is way too much for any algorithm to handle. I guess that in your case, you must really do one of the following:
get 100 computers to cut your parameter testing time to some reasonable time
really try to work out your parameters by hand and mind. There must be some redundancy and at least some sanity check so you can test your case in <1min
for possible result sets, try to figure out some 'operations' that modify it slightly instead of just randomizing it. For example, in TSP some basic operator is lambda, that swaps two nodes and thus creates new route. Your can be shifting some number up/down for some value.
then, find yourself some nice algorithm, your starting point can be somewhere here. The book is invaluable resource for anyone who starts with problem-solving.
How would you mathematically model the distribution of repeated real life performance measurements - "Real life" meaning you are not just looping over the code in question, but it is just a short snippet within a large application running in a typical user scenario?
My experience shows that you usually have a peak around the average execution time that can be modeled adequately with a Gaussian distribution. In addition, there's a "long tail" containing outliers - often with a multiple of the average time. (The behavior is understandable considering the factors contributing to first execution penalty).
My goal is to model aggregate values that reasonably reflect this, and can be calculated from aggregate values (like for the Gaussian, calculate mu and sigma from N, sum of values and sum of squares). In other terms, number of repetitions is unlimited, but memory and calculation requirements should be minimized.
A normal Gaussian distribution can't model the long tail appropriately and will have the average biased strongly even by a very small percentage of outliers.
I am looking for ideas, especially if this has been attempted/analysed before. I've checked various distributions models, and I think I could work out something, but my statistics is rusty and I might end up with an overblown solution. Oh, a complete shrink-wrapped solution would be fine, too ;)
Other aspects / ideas: Sometimes you get "two humps" distributions, which would be acceptable in my scenario with a single mu/sigma covering both, but ideally would be identified separately.
Extrapolating this, another approach would be a "floating probability density calculation" that uses only a limited buffer and adjusts automatically to the range (due to the long tail, bins may not be spaced evenly) - haven't found anything, but with some assumptions about the distribution it should be possible in principle.
Why (since it was asked) -
For a complex process we need to make guarantees such as "only 0.1% of runs exceed a limit of 3 seconds, and the average processing time is 2.8 seconds". The performance of an isolated piece of code can be very different from a normal run-time environment involving varying levels of disk and network access, background services, scheduled events that occur within a day, etc.
This can be solved trivially by accumulating all data. However, to accumulate this data in production, the data produced needs to be limited. For analysis of isolated pieces of code, a gaussian deviation plus first run penalty is ok. That doesn't work anymore for the distributions found above.
[edit] I've already got very good answers (and finally - maybe - some time to work on this). I'm starting a bounty to look for more input / ideas.
Often when you have a random value that can only be positive, a log-normal distribution is a good way to model it. That is, you take the log of each measurement, and assume that is normally distributed.
If you want, you can consider that to have multiple humps, i.e. to be the sum of two normals having different mean. Those are a bit tricky to estimate the parameters of, because you may have to estimate, for each measurement, its probability of belonging to each hump. That may be more than you want to bother with.
Log-normal distributions are very convenient and well-behaved. For example, you don't deal with its average, you deal with it's geometric mean, which is the same as its median.
BTW, in pharmacometric modeling, log-normal distributions are ubiquitous, modeling such things as blood volume, absorption and elimination rates, body mass, etc.
ADDED: If you want what you call a floating distribution, that's called an empirical or non-parametric distribution. To model that, typically you save the measurements in a sorted array. Then it's easy to pick off the percentiles. For example the median is the "middle number". If you have too many measurements to save, you can go to some kind of binning after you have enough measurements to get the general shape.
ADDED: There's an easy way to tell if a distribution is normal (or log-normal). Take the logs of the measurements and put them in a sorted array. Then generate a QQ plot (quantile-quantile). To do that, generate as many normal random numbers as you have samples, and sort them. Then just plot the points, where X is the normal distribution point, and Y is the log-sample point. The results should be a straight line. (A really simple way to generate a normal random number is to just add together 12 uniform random numbers in the range +/- 0.5.)
The problem you describe is called "Distribution Fitting" and has nothing to do with performance measurements, i.e. this is generic problem of fitting suitable distribution to any gathered/measured data sample.
The standard process is something like that:
Guess the best distribution.
Run hypothesis tests to check how well it describes gathered data.
Repeat 1-3 if not well enough.
You can find interesting article describing how this can be done with open-source R software system here. I think especially useful to you may be function fitdistr.
In addition to already given answers consider Empirical Distributions. I have successful experience in using empirical distributions for performance analysis of several distributed systems. The idea is very straightforward. You need to build histogram of performance measurements. Measurements should be discretized with given accuracy. When you have histogram you could do several useful things:
calculate the probability of any given value (you are bound by accuracy only);
build PDF and CDF functions for the performance measurements;
generate sequence of response times according to a distribution. This one is very useful for performance modeling.
Try whit gamma distribution http://en.wikipedia.org/wiki/Gamma_distribution
From wikipedia
The gamma distribution is frequently a probability model for waiting times; for instance, in life testing, the waiting time until death is a random variable that is frequently modeled with a gamma distribution.
The standard for randomized Arrival times for performance modelling is either Exponential distribution or Poisson distribution (which is just the distribution of multiple Exponential distributions added together).
Not exactly answering your question, but relevant still: Mor Harchol-Balter did a very nice analysis of the size of jobs submitted to a scheduler, The effect of heavy-tailed job size distributions on computer systems design (1999). She found that the size of jobs submitted to her distributed task assignment system took a power-law distribution, which meant that certain pieces of conventional wisdom she had assumed in the construction of her task assignment system, most importantly that the jobs should be well load balanced, had awful consequences for submitters of jobs. She's done good follor-up work on this issue.
The broader point is, you need to ask such questions as:
What happens if reasonable-seeming assumptions about the distribution of performance, such as that they take a normal distribution, break down?
Are the data sets I'm looking at really representative of the problem I'm trying to solve?