Slashdot Mirror
IBM Claims Breakthrough Energy-Efficient Algorithm
jitendraharlalka sends news of a claimed algorithmic breakthrough by IBM, though from the scant technical detail provided it's hard to tell exactly how important the development might be. IBM apparently presented its results yesterday at the Society for Industrial and Applied Mathematics conference in Seattle. The breathless press release begins: "IBM Research today unveiled a breakthrough method based on a mathematical algorithm that reduces the computational complexity, costs, and energy usage for analyzing the quality of massive amounts of data by two orders of magnitude. This new method will greatly help enterprises extract and use the data more quickly and efficiently to develop more accurate and predictive models. In a record-breaking experiment, IBM researchers used the fourth most powerful supercomputer in the world... to validate nine terabytes of data... in less than 20 minutes, without compromising accuracy. Ordinarily, using the same system, this would take more than a day. Additionally, the process used just one percent of the energy that would typically be required."
Out of smart people, not greenards
And that color would be blue! Hopefully us mere mortals will be able to benefit from such algorihtms.
I tried to think of a good sig, and this wasn't it.
I guess they stopped using Windows Vista?
After we moved from MySQL to PostgreSQL, we saw similar performance improvements. Then we doubled our performance again when we moved to FreeBSD from Linux. We never expected a few software changes to have such a big impact, but were happy that we could reuse all of our existing hardware.
Sounds like someone found a faster algorithm (maybe just constants), and since energy efficiency is the hot new thing, "faster" is now translated into "saves energy".
Can it organise my porn?
I'm all for energy-efficient algorithms and datacenter but this PR is nothing but green-washing. IBM's algorithm is just faster so it uses less energy. Duh.
Automatically spreading loads across datacenters in multiple locations to take advantage of local environmental conditions so you don't have to use chillers, now that's something.
Nobox: Only simple products.
I'll buy three!
What do they do exactly?
-- There are three kinds of mathematicians: those who can add and those who can't.
Can someone please clarify exactly what they've achieved here? All I hear is that they can somehow sift through large quantities of data much quicker. What kind of data? What are they trying to extract? And for what end?
Why OpalCalc is the best Windows calc
No explanation of how these two orders of magnitude of improvement were achieved. Less space than a Nomad. Lame.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
So what kind of porn was that, and why is this of interest to the enterprise?
If you're running the fourth most powerful supercomputer in the world you are not saving energy. Period.
Get this type of 'efficiency' running on mobile & portable devices and laptops first. Then you can claim some sort of energy related victory.
This would be a real story if it gave implementation details, but it doesn't even tell us what the algorithm does; therefore it's totally worthless. Get this crap off the front page.
Without more information, this really sounds like they just had a horribly-slow-but-at-least-it-works algorithm in the first place and now done some work on making it more efficient. They don't even say what type of processing was being done on the data..
which is totally what she said
...for analyzing the quality of massive amounts of data...
I have an algorithm that does that in O(1):
return "Not the best quality, but pretty good.";
My UID is prime. Hah!
Now you just need the brains. Brains to design the system, brains to drive the investigation, and brains to try to improve the algorithms the system uses. ... Er, but what are we going to do with all the people who just don't "have" the brains?
Mmmm, brains ...
just in case you don't know how marketing looks like. Until there is a technical paper from IBM we could just assume that someone said "I have an idea! Lets use quicksort instead of bubble sort!".
and Business Inteligence software. Things that large corps use to help make decisions (Goldman-Sachs?) and manipulate the banks/markets even faster today so Yea! This is a big deal to corps. Not so big a deal to individuals other then the damn corps can make idiot decisions even faster now.
Mod me up/Mod me down: I wont frown as I've no crown
Implemented for Linux, but analogously applicable to other systems. Running this once should reduce your PC's energy consumption to near zoro:
#!/bin/bash
#
# save-energy.sh
#
# Save enormous amounts of energy, irrespective of cost in lost computation
# must be run as root
Of course, effeciency will be lost if you do anything else with your PC (like turn it back on), but hey, no algorithm is perfect for all use cases.
The Future of Human Evolution: Autonomy
Is there hope for a merge of all virtual fragmented universes into a single universe? We we can explore strange new worlds; seek out new life and new civilizations; boldly go where our avatar hasn't gone before?
Build your own energy sources from scratch. http://otherpower.com/
regularly produce this magnitude of algorithm speedup...
The MAFIAA is a bunch of mindless jerks who will be the first up against the wall when the revolution comes
As a computer engineer, I'm fascinated by the potential improvements in performance.
As a wired citizen, I'm terrified of the additional data-mining capabilities this will provide to our corporate overlords.
We are the 198 proof..
./configure
What was the algorithm? For all I know (having not read TFA), it could be that they replaced bubble sort with quicksort.
Here's a link with actual content on what the algorithm does:
http://www.hpcwire.com/features/IBM-Invents-Short-Cut-to-Assessing-Data-Quality-85427987.html
are they marketing the price of the new algorithm? Peak oil trumps software. If they open source it...never mind.
"Low cost high performance uncertainty quantification", full text available in PDF.
http://portal.acm.org/citation.cfm?id=1645421&coll=GUIDE&dl=GUIDE&CFID=77531079&CFTOKEN=42017699&ret=1#Fulltext
And, here's the abstract:
Uncertainty quantification in risk analysis has become a key
application. In this context, computing the diagonal of in-
verse covariance matrices is of paramount importance. Stan-
dard techniques, that employ matrix factorizations, incur a
cubic cost which quickly becomes intractable with the cur-
rent explosion of data sizes. In this work we reduce this
complexity to quadratic with the synergy of two algorithms
that gracefully complement each other and lead to a radi-
cally different approach. First, we turned to stochastic esti-
mation of the diagonal. This allowed us to cast the problem
as a linear system with a relatively small number of multiple
right hand sides. Second, for this linear system we developed
a novel, mixed precision, iterative refinement scheme, which
uses iterative solvers instead of matrix factorizations. We
demonstrate that the new framework not only achieves the
much needed quadratic cost but in addition offers excellent
opportunities for scaling at massively parallel environments.
We based our implementation on BLAS 3 kernels that en-
sure very high processor performance. We achieved a peak
performance of 730 TFlops on 72 BG/P racks, with a sus-
tained performance 73% of theoretical peak. We stress that
the techniques presented in this work are quite general and
applicable to several other important applications.
20 minutes is roughly 1% of "More than a day",
so it's not only using "1% of the energy required",
it's just using 1% of the time required,
so this is not a breakthrough of energy efficiency, it's a "CPU time" breakthrough, LOL
Burn 'em.
Obviously are not releasing details until the Patent application goes through and the Patent Troll company set up. They certainly would not release the information so other people could just steal their idea. Maybe they will package it in sealed application and rent it out. Hmmmm anyone remember the Chess Playing Mechanical Turk? (1770).
From TFS:
"reduces the computational complexity[...]by two orders of magnitude[...]
Additionally, the process used just one percent of the energy that would typically be required"
Well, duh, what's so shocking about a computation taking 1% of the time previously needed now only takes 1% of the energy as well?
...and it automatically fired three project managers.
The description of this "new algorithm" is pretty sparse.
Any word on if it allows faster solutions to encryption problems so that we now all need longer passwords?
I shouldn't be telling anyone this but I found the algorithm. Here it is in Java, please convert it to your preferred language.
public int computeData(Data data) { return 42; }
Oh crap, I forgot to post this anonymously. Why are there mice all over the pl@#M *^&I RCS$WE^%
[CARRIER LOST]
Well, there's spam egg sausage and spam, that's not got much spam in it.
Smart brains ...
Well you could be up here in Canada, where my wife and I are paying about $90 a month for our health care insurance. That would be horrible wouldn't it? Living under an inefficient Socialist healthcare system and saving all that money while getting very good healthcare services?
Now if my income was lower, I would pay less. If it fell below a threshhold I can't recall, it would be free. If I earn more my payments will go up but not onerously so.
For some examples:
1) My mother died of cancer 2 years ago. She received full health care, hospital time, radiation treatment etc. Total cost: $50 for the ambulance that took her to the hospital the last time she went.
2) A friend of mine was diagnosed with a brain tumour. He was in and out of the hospital inside of a few weeks, perfectly healthy after the tumour was removed. Total cost to him: nothing beyond his payments.
3) I am 50 years old, I have been to the doctors all my life. I have never had to pay a dime for a visit. As a result, when I am sick I can go to the doctors, not worry about whether I can afford to find out if its something serious etc.
I know we have some problems from time to time up here, we get just as horrified by mistakes in the system as the anti-healthcare folks who post all those stories do down in the US when they are denegrating our system up here. However, overall it works very well.
Oh its worth pointing out that anyone who wanted different health care, say if my friend wanted treatment for his brain tumour down in the US, is capable of getting that treatment instead - they just have to pay for it.
The system benefits those at the bottom of the economic scale, it doesn't penalize those at the top of the scale
Nothing new, probably just using caching and a makefile-like dependency-checking algorithm.
Kind like you can speed up the calculations in the game of life if you calculate the sums of 3 adjacent cells over x first, and calculate the sums over y of the previous sums. So you use 4 addition operations instead of 8 (and some array indexing operations which I'm too lazy to calculate, probably 6 instead of 8).
Really sad that they will probably patent it so no one else can use it for 20 years.
Hey don't blame me, IANAB
"Er, but what are we going to do with all the people who just don't "have" the brains? They get a free ride?"
This question was already answered by Saturday Morning Breakfast Cereal:
http://www.smbc-comics.com/index.php?db=comics&id=1760#comic
That sounds nice but let's focus cost. If you're only paying $90/month, there's no way you are paying the full cost of health care. According to Wikipedia, the Canadian government pays about 70% of health care, so your $90 represents about $300 in actual cost. (Still a lot better than the US.)
According to Wikipedia, Canada has 6% of the population above 65 years of age. The US has 12.8%.
Eh, what's the point. It's very difficult to compare costs between different countries. I don't even know whether your $90 is converted to US dollars or still in Canadian dollars (hey 5% makes a difference).
I know the US has high health care costs, but the important question is why. People like you don't even pretend to analyze the situation, you just think it's "the system" that makes it cheaper. Give me some facts if you have them! I look for them but it's hard for one person to become an expert, or even just become knowledgeable, about the health care systems of a dozen different countries.
Low Cost High Performance Uncertainty Quantification
http://portal.acm.org/citation.cfm?id=1645421&jmp=cit&coll=portal&dl=ACM&CFID=62671076&CFTOKEN=92670385#CIT
Keywords:
Inverse Covariance Matrices, Stochastic Estimation, Iterative
Renement, Iterative Solvers, Quadratic Cost, Massive
Parallelism
All rites reversed 2010
You managed to optimize your SQL query.
When I optimize my queries, why don't I get press releases?!
"in less than 20 minutes, without compromising accuracy. Ordinarily, using the same system, this would take more than a day"
This just means that they were doing a pretty damn lousy job before they fixed the problem.
The question of whether a computer can think is no more interesting than the question of whether a submarine can swim.
You're right that spinning this as a green result is trendy. But it can't be just a change in constants, since they say they reduced the complexity. "Reducing the complexity" is a technical term that means the speed, as a function of input size n (or l,m,n, etc.), keeps improving as n gets bigger. That is,
lim time_required_new(n) / time_required_old(n) = 0.
n --> infinity
An example is that an O(n log n) algorithm like Heapsort has reduced complexity compared to an O(n**2) algorithm like selection sort. But you can't legitimately claim to have "reduced the complexity" just by cleaning up the code or cutting out a constant factor of work (like going from bubble sort to selection sort).
$META_SIG_JOKE
For anyone who's interested in what these guys have done- the WHPCF'09 paper by Bekas and Fedulova (and going back a bit further, their 2007 paper by Bekas et al.) give the details.
In many statistical problems we end up with the problem of finding the diagonal entries of the inverse of a known symmetric and positive definite matrix A. For example, in linear regression the variances for the fitted parameters are found on the diagonal of inv(X'*X). When this matrix A is very large, the computation can be very expensive, since it requires O(N^3) time by conventional methods (Compute the Cholesky factorization of A and then use the Cholesky factors to solve for N right hand sides.)
Bekas et al. have developed a Monte Carlo approach that can give good (e.g. 2-3 digits of accuracy) estimates of the diagonal entries in inv(A) by using an interative method to approximately solve systems of linear equations involving A. The approximate iterative solutions take roughly O(N^2) time, and there are s of these systems to solve, where sN. Thus the computational complexity is lowered from O(N^3) to roughly O(N^2). Furthermore, you can solve these s systems of equations in parallel. Going one step further, you can do a lot of the computation in single precision, so it can be done on GPGPU's and other machines that don't do double precision floating point efficiently.
So imagine if you had a Beowulf cluster of these, they could like, self-power forever?
This likely involves cats strapped to buttered bread, for maximum effect.