S.o.s!!

[Sip]

Re: S.o.s!!

HA! I found some of my old notes from his talk and the conversation I had with him afterwards!!! ... and I almost never take notes. Karoaper, if you are interested, I can email you more details.

[Sip]

Re: S.o.s!!

There are countless examples of this. By the way, here's an interesting fact: the linear programming problem was first shown to be solvable in polynomial time by Leonid Khachiyan in 1979.

I had the pleasure of attending a talk by Andranik Mirzaian a few years ago on this topic and I was absolutely blown away by the elegance of his solution to LP.

I will admit that I never quite understood how exactly Khachiyan's ellipsoid method works but from what I understand, the complexity there is still exponential in terms of the bit size of data (i.e. it's weakly polynomial). As the data can get quite large, this is not really that good in practice and that's why the old Simplex methods have been working really well since Danzig's days. Mirzaian was claiming his algorithm is strongly polynomial (polynomial even on the bit size of the data) which of course is a very amazing claim. Even though his approach seemed very cool at the time, he never produced a proof and I can't find any publications on the topic.

I really hope he has the result and is only sitting on it to polish it up before publishing!!! It would be a shame if he found some flaw in it ... but if it is in fact still a valid result, this would be a REALLY cool thing as Armenians will pretty much OWN LP

[Sip]

Re: S.o.s!!

So, I would say, whether algorithms are relevant or not, depends on which type of an algorithm is in question.

AMEN to that. I have tried so hard to get this across to Siamanto but he is just stuck in his tunnel vision. Normally I would be much more polite in my replies but that strategy just doesn't seem to work with him. He keeps mentioning large enterprises, code reuse, modeling, large scale architectures (software) but fails to recognize that all of what he is talking about is restricted to a very specific and limited computing domain and is not at all universal.

There are many examples of other domains of computing where his view of the world simply do not apply. I will give a simple example of web search ... you type a query and get a result in fraction of a second. Now think what went on for you to get that result... How dynamic the data is ... how large the data is ... how limited the query is and how specific the results have to be ... and potentially just how many billions of such requests you have to process on a daily basis! If after thinking about this seriously for a few minutes you still want to claim working on such an algorithm is not a challenge, then either you are an ultimate genius waaaaaaaay beyond me in your capabilities, or you are an idiot.

[Sip]

Re: S.o.s!!

Siamanto, you are CLEARLY demonstrating how little knowledge you have of how algorithms have been evolving. You see the word "algorithm" and you seem to associate it with known results such as QuickSort or Dijkstra's shortest path or some other elementary text book example. Just making a RIDICULOUS statement such as "algorithms are not a challenge anymore" (which I do paraphrase to "dismissing as child's play" whether you like it or not) shows you have a LOT of catching up to do in this domain.

If you really think algorithms are not a challenge anymore, I guess you are quite satisfied with your email clients ability to filter spam, the amazing efficiency of airlines, traffic congestions in large cities, the search results you get from google, etc etc etc ...

Do you REALLY want to insist on claiming that those are solved problems? I am not even going to bring AI into this ... as I'm sure you will claim that AI algorithms are also not a challenge and they are good enough to build your airport.

[Siamanto]

Re: S.o.s!!

In the 11th hour of this discussion, I thought I'd put my two "kopeks" in.

In my mind, computer science has the two key aspects: the theory, which is mathematical in nature, and implementation, which is based on writing an efficient algorithm that implements the developed theory. You can have a beautiful theory, but in computer science, developing an efficient and correct algorithm for the theory is almost as important.

As an example, different rich machine learning theories and principles would not be as relevant without clever and robust algorithms that realize these principles. As such, these algorithms can be and indeed are very complex. And a computer scientist's job is not only to devise theories but also develop tractable algorithms. In certain cases, algorithm construction is a theory in itself.

There are countless examples of this. By the way, here's an interesting fact: the linear programming problem was first shown to be solvable in polynomial time by Leonid Khachiyan in 1979.

So, I would say, whether algorithms are relevant or not, depends on which type of an algorithm is in question. I'm currently working on an algorithm for a machine learning classifier idea I have thought up. Making the algorithm outperform the competition speedwise is big challenge.

karoaper,

The issue is not theory vs. implementation - by the way, even when it comes to Algorithmics one can distinguish theory and implementation.
Also, I have never said that algorithms are not relevant. As I've told Sip, in my previous post:
"I wonder how many times a simple statement should be repeated...: Nobody is "dismissing algorithms as child's play" or as irrelevant; however, they are not challenges anymore."

Lacking time, I will be as succint as possible!

The point is that, though algorithms are used in order to implement solutions to business problems, they are fairly well known/understood and are not challenges anymore; however, the following - to name only a few - are far more crucial challenges. They are not theoretical issues; those are real problems that enterprises face on a daily basis and their solutions include myriad of different algorithms. Each of the following present many challenges and it is beyond the scope of my reply to even enumerate them.

1- Repeatability in software production: of course, it is a challenge when many hundreds are involved in the process and the different teams - client, managers, business analysts, architects, development teams, QA technicians, admins etc. - are geographically distributed. It's also about scale. This is not about "college level" projects where small teams are involved, having an almost flat organizational structure and almost total lack of process. Also, college projects are not deployed on thousands of sites, not used by thousands of users and not maintained for the next decade or so by hundreds of different developers

2- Reusability of software artifacts in order to improve quality, reduce costs, manage risks...

3- Data Integration and sharing between business partners and within different departments of an enterprise

4- Integration of components developed by different teams into complex systems as solutions to business problems; solutions that can gracefully be extended, maintained and evolve

5- Performance, Scalability i.e. millions of transactions and huge grids where failure is not an option. Keep in mind that the number of transactions and processing are increasing exponentially and solutions are expected to stay operational for at least 10+ years

6- ....

In a different thread, I used the metaphor of a shack - that can be built by a small team or an individual - and high traffic modern airport - that requires the collaboration of many organizations and involve many different technologies. A shack can be "hacked" with simple tools and algorithms; a high traffic airport can only be built with a completely different set of tools, technologies, standards, practices, organization, process, methodologies etc. that include a myriad of algorithms, but the algorithms are not the challenge.
As I see it, Sip does not see beyond the challenges of building a shack. A high traffic airport is a reality, not theory.

[karoaper]

Re: S.o.s!!

In the 11th hour of this discussion, I thought I'd put my two "kopeks" in.

In my mind, computer science has the two key aspects: the theory, which is mathematical in nature, and implementation, which is based on writing an efficient algorithm that implements the developed theory. You can have a beautiful theory, but in computer science, developing an efficient and correct algorithm for the theory is almost as important.

As an example, different rich machine learning theories and principles would not be as relevant without clever and robust algorithms that realize these principles. As such, these algorithms can be and indeed are very complex. And a computer scientist's job is not only to devise theories but also develop tractable algorithms. In certain cases, algorithm construction is a theory in itself.

There are countless examples of this. By the way, here's an interesting fact: the linear programming problem was first shown to be solvable in polynomial time by Leonid Khachiyan in 1979.

So, I would say, whether algorithms are relevant or not, depends on which type of an algorithm is in question. I'm currently working on an algorithm for a machine learning classifier idea I have thought up. Making the algorithm outperform the competition speedwise is big challenge.

[Siamanto]

Re: S.o.s!!

I would suggest a look at this article:

http://economist.com/business/displaystory.cfm?story_id=9795140

It is a fairly light read targeted at a broad audience so hopefully it won't be too confusing for your highness. It might give you an idea why you shouldn't be so quick to dismiss algorithms as child's play if you consider yourself a serious computer scientist who is willing to keep up with the times and not assume what he or she learned "back in college" is still relevant.

I wonder how many times a simple statement should be repeated before you finally grasp it: Nobody is "dismissing algorithms as child's play" or as irrelevant; however, they are not challenges anymore.

I hope that someday you will get out of the hole that has been Academia during the last two decades - I mean in Computer Science - and be exposed to real "enterprise level" projects. Actually, during the last decades, CS evolved thanks to the non-Academic organizations and efforts; while the contribution of Academia has been minor, if not negligible.

Let's see how many more times the above simple statement will have to be repeated???

[Sip]

Re: S.o.s!!

Yes, a dream come true to discuss algorithms and Programming on a forum....maybe, when I was in college????
I would have considered exchanging ideas if it was about more evolved, current and challenging domains, aspects or issues of IT/CS.

I would suggest a look at this article:

http://economist.com/business/displaystory.cfm?story_id=9795140

It is a fairly light read targeted at a broad audience so hopefully it won't be too confusing for your highness. It might give you an idea why you shouldn't be so quick to dismiss algorithms as child's play if you consider yourself a serious computer scientist who is willing to keep up with the times and not assume what he or she learned "back in college" is still relevant.

[Sip]

Re: S.o.s!!

Siamanto, at least you have been the most entertaining thing to hit this forum in the past couple of years.

[Siamanto]

Re: S.o.s!!

I must hand it to Siamanto. No one holds grudges like him/her

More of your cheap calumniation??? At least, they're foolish enough to sound like jokes and amuse me.

Considering that you have systematically brought up the past with such obsession all along in this thread, while I ignored them, do you honestly think your emotional obsessions are not obvious enough for the forum to see? Please keep in mind that it's really hard to miss.