I have a project where I am asked to develop an application to simulate how different page replacement algorithms perform (with varying working set size and stability period). My results:
Vertical axis: page faults
Horizontal axis: working set size
Depth axis: stable period
Are my results reasonable? I expected LRU to have better results than FIFO. Here, they are approximately the same.
For random, stability period and working set size doesnt seem to affect the performance at all? I expected similar graphs as FIFO & LRU just worst performance? If the reference string is highly stable (little branches) and have a small working set size, it should still have less page faults that an application with many branches and big working set size?
More Info
My Python Code | The Project Question
Length of reference string (RS): 200,000
Size of virtual memory (P): 1000
Size of main memory (F): 100
number of time page referenced (m): 100
Size of working set (e): 2 - 100
Stability (t): 0 - 1
Working set size (e) & stable period (t) affects how reference string are generated.
|-----------|--------|------------------------------------|
0 p p+e P-1
So assume the above the the virtual memory of size P. To generate reference strings, the following algorithm is used:
Repeat until reference string generated
pick m numbers in [p, p+e]. m simulates or refers to number of times page is referenced
pick random number, 0 <= r < 1
if r < t
generate new p
else (++p)%P
UPDATE (In response to #MrGomez's answer)
However, recall how you seeded your input data: using random.random,
thus giving you a uniform distribution of data with your controllable
level of entropy. Because of this, all values are equally likely to
occur, and because you've constructed this in floating point space,
recurrences are highly improbable.
I am using random, but it is not totally random either, references are generated with some locality though the use of working set size and number page referenced parameters?
I tried increasing the numPageReferenced relative with numFrames in hope that it will reference a page currently in memory more, thus showing the performance benefit of LRU over FIFO, but that didn't give me a clear result tho. Just FYI, I tried the same app with the following parameters (Pages/Frames ratio is still kept the same, I reduced the size of data to make things faster).
--numReferences 1000 --numPages 100 --numFrames 10 --numPageReferenced 20
The result is
Still not such a big difference. Am I right to say if I increase numPageReferenced relative to numFrames, LRU should have a better performance as it is referencing pages in memory more? Or perhaps I am mis-understanding something?
For random, I am thinking along the lines of:
Suppose theres high stability and small working set. It means that the pages referenced are very likely to be in memory. So the need for the page replacement algorithm to run is lower?
Hmm maybe I got to think about this more :)
UPDATE: Trashing less obvious on lower stablity
Here, I am trying to show the trashing as working set size exceeds the number of frames (100) in memory. However, notice thrashing appears less obvious with lower stability (high t), why might that be? Is the explanation that as stability becomes low, page faults approaches maximum thus it does not matter as much what the working set size is?
These results are reasonable given your current implementation. The rationale behind that, however, bears some discussion.
When considering algorithms in general, it's most important to consider the properties of the algorithms currently under inspection. Specifically, note their corner cases and best and worst case conditions. You're probably already familiar with this terse method of evaluation, so this is mostly for the benefit of those reading here whom may not have an algorithmic background.
Let's break your question down by algorithm and explore their component properties in context:
FIFO shows an increase in page faults as the size of your working set (length axis) increases.
This is correct behavior, consistent with Bélády's anomaly for FIFO replacement. As the size of your working page set increases, the number of page faults should also increase.
FIFO shows an increase in page faults as system stability (1 - depth axis) decreases.
Noting your algorithm for seeding stability (if random.random() < stability), your results become less stable as stability (S) approaches 1. As you sharply increase the entropy in your data, the number of page faults, too, sharply increases and propagates the Bélády's anomaly.
So far, so good.
LRU shows consistency with FIFO. Why?
Note your seeding algorithm. Standard LRU is most optimal when you have paging requests that are structured to smaller operational frames. For ordered, predictable lookups, it improves upon FIFO by aging off results that no longer exist in the current execution frame, which is a very useful property for staged execution and encapsulated, modal operation. Again, so far, so good.
However, recall how you seeded your input data: using random.random, thus giving you a uniform distribution of data with your controllable level of entropy. Because of this, all values are equally likely to occur, and because you've constructed this in floating point space, recurrences are highly improbable.
As a result, your LRU is perceiving each element to occur a small number of times, then to be completely discarded when the next value was calculated. It thus correctly pages each value as it falls out of the window, giving you performance exactly comparable to FIFO. If your system properly accounted for recurrence or a compressed character space, you would see markedly different results.
For random, stability period and working set size doesn't seem to affect the performance at all. Why are we seeing this scribble all over the graph instead of giving us a relatively smooth manifold?
In the case of a random paging scheme, you age off each entry stochastically. Purportedly, this should give us some form of a manifold bound to the entropy and size of our working set... right?
Or should it? For each set of entries, you randomly assign a subset to page out as a function of time. This should give relatively even paging performance, regardless of stability and regardless of your working set, as long as your access profile is again uniformly random.
So, based on the conditions you are checking, this is entirely correct behavior consistent with what we'd expect. You get an even paging performance that doesn't degrade with other factors (but, conversely, isn't improved by them) that's suitable for high load, efficient operation. Not bad, just not what you might intuitively expect.
So, in a nutshell, that's the breakdown as your project is currently implemented.
As an exercise in further exploring the properties of these algorithms in the context of different dispositions and distributions of input data, I highly recommend digging into scipy.stats to see what, for example, a Gaussian or logistic distribution might do to each graph. Then, I would come back to the documented expectations of each algorithm and draft cases where each is uniquely most and least appropriate.
All in all, I think your teacher will be proud. :)
Related
I have a problem that I don't have enough training data for my NN. It is trying to predict the result of a soccer game given the last games which I woulf say is a regression task.
The training data are results of soccer games of the last 15 seasons (which are about 4500 games). Getting to new data would be hard and would take a lot of time.
What should I do now?
Is it good to duplicate the data?
Should I input randomized data? (Maybe noise but I'm not quite sure what that is)
If there is no way of creating more data,
I should probably turn up the learning rate right? (I have it sitting at 0.01 and the momentum at 0.9)
I am using mini batches consisting of 32 training datas in training. Since I don't have a lot of training I don't have a lot of mini batches. Should I stop using them?
To start from the beginning: This is a very theoretical question and is not directly related to programming, which I recommend (in future) to post over at the Data Science Stackexchange.
To go into your problem: 4500 samples is not as bad as it sounds, depending on the exact task at hand. Are you trying to predict the match results (i.e. which team is the winner?), are you looking for more specific predictions (across a lot of different, specific teams)?
If you can make sure that you have a reasonable amount of data per class, one can work with a number of samples lower than what you have. Simply duplicating the data will not help you much, since you are very likely to just overfit on the samples you are seeing, without much of an improvement; Or rather, you will get the same results as training over a longer period (since essentially you see every sample twice per epoch, instead of one).
Again, what usually happens after long training periods is overfitting, so nothing gained here.
Your second suggestion is generally called data augmentation. Instead of simply copying samples, you alter them enough to make it look "different" to the network. But be careful! Data augmentation works well for some inputs, like images, since the change in input is significant enough to not represent the same sample, but still contains meaningful information about the class (a horizontally mirrored image of a cat still shows a "valid cat", unlike a vertically mirrored image, which is more unrealistic in the real world).
Essentially, it depends on your input features to determine where it makes sense to add noise. If you are only changing the results of the previous game, a minor change in input (adding/subtracting one goal at random) can significantly change the prediction you make.
If you slightly scramble ELO scores by a random number, on the other hand, the input value will not be too different, "but different enough" to use it as a novel example.
Turning up the learning rate is not a good idea, since you are essentially just letting the network converge more towards the specific samples. On the contrary, I would argue that the current learning rate is still too high, and you should certainly not increase it.
Regarding mini batches, I think I have referenced this a million times now, but always consider smaller minibatches. From a theoretical point of view, you are more likely to converge to a local minimum.
In the documentation of H2O is written:
mini_batch_size: Specify a value for the mini-batch size. (Smaller values lead to a better fit; larger values can speed up and generalize better.)
but when I run a model using the FLOW UI (with mini_batch_size > 1) in the log file is written:
WARN: _mini_batch_size Only mini-batch size = 1 is supported right now.
so the question: is the mini_batch_size really used??
It appears to be a left over from preparation for a DeepWater integration that never happened. E.g. https://github.com/h2oai/h2o-3/search?l=Java&p=2&q=mini_batch_size
That makes sense, because the Hogwild! algorithm, that H2O's deep learning uses, does away with the need for batching training data.
To sum up, I don't think it is used.
Using Matlab, I am going to generate several data files and store them in H5 format as 20x1500xN, where N is an integer that can vary, but typically around 2300. Each file will have 4 different data sets with equal structure. Thus, I will quickly achieve a storage problem. My two questions:
Is there any reason not the split the 4 different data sets, and just save as 4x20x1500xNinstead? I would prefer having them split, since it is different signal modalities, but if there is any computational/compression advantage to not having them separated, I will join them.
Using Matlab's built-in compression, I set deflate=9 (and DataType=single). However, I have now realized that using deflate multiplies my computational time with 5. I realize this could have something to do with my ChunkSize, which I just put to 20x1500x5 - without any reasoning behind it. Is there a strategic way to optimize computational load w.r.t. deflation and compression time?
Thank you.
1- Splitting or merging? It won't make a difference in the compression procedure, since it is performed in blocks.
2- Your choice of chunkshape seems, indeed, bad. Chunksize determines the shape and size of each block that will be compressed independently. The bad is that each chunk is of 600 kB, that is much larger than the L2 cache, so your CPU is likely twiddling its fingers, waiting for data to come in. Depending on the nature of your data and the usage pattern you will use the most (read the whole array at once, random reads, sequential reads...) you may want to target the L1 or L2 sizes, or something in between. Here are some experiments done with a Python library that may serve you as a guide.
Once you have selected your chunksize (how many bytes will your compression blocks have), you have to choose a chunkshape. I'd recommend the shape that most closely fits your reading pattern, if you are doing partial reads, or filling in in a fastest-axis-first if you want to read the whole array at once. In your case, this will be something like 1x1500x10, I think (second axis being the fastest, last one the second fastest, and fist the slowest, change if I am mistaken).
Lastly, keep in mind that the details are quite dependant on the specific machine you run it: the CPU, the quality and load of the hard drive or SSD, speed of RAM... so the fine tuning will always require some experimentation.
I am using a startech capture card for capturing video from the source machine..I have encoded that video using matlab so every frame of that video will contain that marker...I run that video on the source computer(HDMI out) connected via HDMI to my computer(HDMI IN) once i capture the frame as bitmap(1920*1080) i re-size it to 1280*720 i send it for processing , the processing code checks every pixel for that marker.
The issue is my capture card is able to capture only at 1920*1080 where as the video is of 1280*720. Hence in order to retain the marker I am down scaling the frame captured to 1280*720 which in turn alters the entire pixel array I believe and hence I am not able to retain marker I fed in to the video.
In that capturing process the image is going through up-scaling which in turn changes the pixel values.
I am going through few research papers on Steganography but it hasn't helped so far. Is there any technique that could survive image resizing and I could retain pixel values.
Any suggestions or pointers will be really appreciated.
My advice is to start with searching for an alternative software that doesn't rescale, compress or otherwise modify any extracted frames before handing them to your control. It may save you many headaches and days worth of time. If you insist on implementing, or are forced to implement a steganography algorithm that survives resizing, keep on reading.
I can't provide a specific solution because there are many ways this can be (possibly) achieved and they are complex. However, I'll describe the ingredients a solution will most likely involve and your limitations with such an approach.
Resizing a cover image is considered an attack as an attempt to destroy the secret. Other such examples include lossy compression, noise, cropping, rotation and smoothing. Robust steganography is the medicine for that, but it isn't all powerful; it may be able to provide resistance to only specific types attacks and/or only small scale attacks at that. You need to find or design an algorithm that suits your needs.
For example, let's take a simple pixel lsb substitution algorithm. It modifies the lsb of a pixel to be the same as the bit you want to embed. Now consider an attack where someone randomly applies a pixel change of -1 25% of the time, 0 50% of the time and +1 25% of the time. Effectively, half of the time it will flip your embedded bit, but you don't know which ones are affected. This makes extraction impossible. However, you can alter your embedding algorithm to be resistant against this type of attack. You know the absolute value of the maximum change is 1. If you embed your secret bit, s, in the 3rd lsb, along with setting the last 2 lsbs to 01, you guarantee to survive the attack. More specifically, you get xxxxxs01 in binary for 8 bits.
Let's examine what we have sacrificed in order to survive such an attack. Assuming our embedding bit and the lsbs that can be modified all have uniform probabilities, the probability of changing the original pixel value with the simple algorithm is
change | probability
-------+------------
0 | 1/2
1 | 1/2
and with the more robust algorithm
change | probability
-------+------------
0 | 1/8
1 | 1/4
2 | 3/16
3 | 1/8
4 | 1/8
5 | 1/8
6 | 1/16
That's going to affect our PSNR quite a bit if we embed a lot of information. But we can do a bit better than that if we employ the optimal pixel adjustment method. This algorithm minimises the Euclidean distance between the original value and the modified one. In simpler terms, it minimises the absolute difference. For example, assume you have a pixel with binary value xxxx0111 and you want to embed a 0. This means you have to make the last 3 lsbs 001. With a naive substitution, you get xxxx0001, which has a distance of 6 from the original value. But xxx1001 has only 2.
Now, let's assume that the attack can induce a change of 0 33.3% of the time, 1 33.3% of the time and 2 33.3%. Of that last 33.3%, half the time it will be -2 and the other half it will be +2. The algorithm we described above can actually survive a +2 modification, but not a -2. So 16.6% of the time our embedded bit will be flipped. But now we introduce error correcting codes. If we apply such a code that has the potential to correct on average 1 error every 6 bits, we are capable of successfully extracting our secret despite the attack altering it.
Error correction generally works by adding some sort of redundancy. So even if part of our bit stream is destroyed, we can refer to that redundancy to retrieve the original information. Naturally, the more redundancy you add, the better the error correction rate, but you may have to double the redundancy just to improve the correction rate by a few percent (just arbitrary numbers here).
Let's appreciate here how much information you can hide in a 1280x720 (grayscale) image. 1 bit per pixel, for 8 bits per letter, for ~5 letters per word and you can hide 20k words. That's a respectable portion of an average novel. It's enough to hide your stellar Masters dissertation, which you even published, in your graduation photo. But with a 4 bit redundancy per 1 bit of actual information, you're only looking at hiding that boring essay you wrote once, which didn't even get the best mark in the class.
There are other ways you can embed your information. For example, specific methods in the frequency domain can be more resistant to pixel modifications. The downside of such methods are an increased complexity in coding the algorithm and reduced hiding capacity. That's because some frequency coefficients are resistant to changes but make embedding modifications easily detectable, then there are those that are fragile to changes but they are hard to detect and some lie in the middle of all of this. So you compromise and use only a fraction of the available coefficients. Popular frequency transforms used in steganography are the Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT).
In summary, if you want a robust algorithm, the consistent themes that emerge are sacrificing capacity and applying stronger distortions to your cover medium. There have been quite a few studies done on robust steganography for watermarks. That's because you want your watermark to survive any attacks so you can prove ownership of the content and watermarks tend to be very small, e.g. a 64x64 binary image icon (that's only 4096 bits). Even then, some algorithms are robust enough to recover the watermark almost intact, say 70-90%, so that it's still comparable to the original watermark. In some case, this is considered good enough. You'd require an even more robust algorithm (bigger sacrifices) if you want a lossless retrieval of your secret data 100% of the time.
If you want such an algorithm, you want to comb the literature for one and test any possible candidates to see if they meet your needs. But don't expect anything that takes only 15 lines to code and 10 minutes of reading to understand. Here is a paper that looks like a good start: Mali et al. (2012). Robust and secured image-adaptive data hiding. Digital Signal Processing, 22(2), 314-323. Unfortunately, the paper is not open domain and you will either need a subscription, or academic access in order to read it. But then again, that's true for most of the papers out there. You said you've read some papers already and in previous questions you've stated you're working on a college project, so access for you may be likely.
For this specific paper, table 4 shows the results of resisting a resizing attack and section 4.4 discusses the results. They don't explicitly state 100% recovery, but only a faithful reproduction. Also notice that the attacks have been of the scale 5-20% resizing and that only allows for a few thousand embedding bits. Finally, the resizing method (nearest neighbour, cubic, etc) matters a lot in surviving the attack.
I have designed and implemented ChromaShift: https://www.facebook.com/ChromaShift/
If done right, steganography can resiliently (i.e. robustly) encode identifying information (e.g. downloader user id) in the image medium while keeping it essentially perceptually unmodified. Compared to watermarks, steganography is a subtler yet more powerful way of encoding information in images.
The information is dynamically multiplexed into the Cb Cr fabric of the JPEG by chroma-shifting pixels to a configurable small bump value. As the human eye is more sensitive to luminance changes than to chrominance changes, chroma-shifting is virtually imperceptible while providing a way to encode arbitrary information in the image. The ChromaShift engine does both watermarking and pure steganography. Both DRM subsystems are configurable via a rich set of of options.
The solution is developed in C, for the Linux platform, and uses SWIG to compile into a PHP loadable module. It can therefore be accessed by PHP scripts while providing the speed of a natively compiled program.
My question is specific to iPhone, iPod, and iPad, since I am assuming that the architecture makes a big difference. I'm hoping there is either a specification somewhere (for the various chips perhaps), or a reliable way to measure T for each specific instruction. I know I can use any number of tools to measure aggregate processor time used, memory used, etc. I want to quantify at a lower level.
So, I'm able to figure out how many times I go through the main part of the algorithm. For example, I iterate n * (n-1) times in a naive implementation, and between n (best case) and n + n * (n-1) (worst case) in another. I can also make a reasonable count of the total number of instructions (+ - = % * /, and logic statements), and I can compare those counts, but that's assuming the weight of each operation is the same. Also, I don't have any idea how to weight the actual time value of a logic statement (if, else, for, while) vs a mathematical operator... is "if" as much work as "+" each time I use it? I would love to know where to find this information.
So, for clarity, my goal is to discover how much processor time I am demanding of the CPU (or GPU or any U) so that I can design an optimal algorithm around processor time. Can someone give me an idea of where to start for iOS hardware?
Edit: This link to ClockServices.c and SIMD stuff in the developer portal might be a good start for people interested in this. A few more cups of coffee tonight and I might get through it ;)
On a modern platform, processor time isn't the only limiting factor. Often, memory access is.
Still, processor time:
Your basic approach at an estimation for the processor load is OK, though, and is sensible: Make a rough estimate of the cost based on your knowledge of typical platforms.
In this article, Table 1 shows the times for typical primitive operations in .NET. While your platform may vary, the relative time is usually very similar. Maybe you can find - or even make - one for iStuff.
(I haven't come across one so thorough for other platforms, except processor / instruction set manuals, but they deal with assembly instructions)
memory locality:
A cache miss can cost you hundreds of cycles, a disk access a thousand times as much. So controlling your memory access patterns (i.e. reducing the working set, restructuring and accessing data in a cache-friendly way) is an important part of evaluating an algorithm.
xCode has instruments to measure performance of each function/operation, you can simply use them.