S.o.s!!

[Siamanto]

Re: S.o.s!!

It's pointless to keep on opening new subjects when already open ones remain unanswered... As I have promised, I'll address the new ones once the already open ones are closed.

You always wanted to be given the chance to prove yourself - i.e. your silly "contest" or whatever, so here's the opportunity. Instead of conveniently hiding, why don't you?
1- Functionally specify the problem
2- Describe a solution: architecture and technology
3- Show that the challenge is algorithmic in nature

Siamanto, I really don't think I am capable of answering your questions.

Thank you for being honest, honesty adds clarity, transparency and trust to the interaction and promotes respect; now, we can consider the optimization problems that you have described. A couple of preliminary comments/questions:

1. From a CS/IT perspective, I see both problems as similar and whatever I say about one applies to both or any other constraints based optimization problem - regardless of the application domain

2. Considering the two most important phases of such an "optimization project" i.e.
2.1 Design of the Mathematical Model and all relevant heuristics
2.2 Implementation - i.e. coding - of the outcome of the design phase
Was the scope of your question the design phase, the implementation or both?

Also, sorry about all the bad stuff I have said about you and I hope you will forgive me one day.

I have nothing to forgive to anybody - it's not how I function; if someone "pushes" me - or pushes it - than I push the person back; otherwise, I do my best to respect the person. In other words, it's up to you, not up to me.

When walking on a sidewalk, I stop to give a frightened pigeon, running away from a rushing passerby, the priority; however, I may confront a person who seems to be pushing or being dishonest with or taking advantage of or manipulate me or, in some cases, others - no matter how "powerful" the person seems. I'm very comfortable with myself.

In any case, to be honest, I ignore the details when your monologue is too irrelevant to my post, and my self-image does not depend on what some may perceive me or tell me.

[Siamanto]

Re: S.o.s!!

now, we can consider the optimization problems that you have described. A couple of preliminary comments/questions:

1. From a CS/IT perspective, I see both problems as similar and whatever I say about one applies to both or any other constraints based optimization problem - regardless of the application domain

2. Considering the two most important phases of such an "optimization project" i.e.
2.1 Design of the Mathematical Model and all relevant heuristics
2.2 Implementation - i.e. coding - of the outcome of the design phase
Was the scope of your question the design phase, the implementation or both?

The following may not be polished enough but it's good enough for our purposes; in any case, is a good starting point.

Design Phase
There may be a different way of organizing the activities of the Design Phase; but, the following seems appropriate enough for our purposes:
1. Design of the Mathematical Model and all relevant heuristics, simplifications...
2. Design issues relevant to the implementation of the output of Task 1
(Of course, depending on the CS/IT culture and/or considering the relative simplicity of the system, some may either skip or include those activities as part of the Implementation Phase.)

The activities of Task1 can be ignored because they are of no relevance to CS/IT and are Mathematical in nature.

The relevant questions that Task2 should answer are
1. Based on the specifications, how configurable and reusable should be the solution - i.e. as reusable as a product or should it be highly customized to the specific application?
In our case, the answer to the above would influence the following
1.1 Should the system leverage a database to store the characteristics of the elements - i.e. boxes, transistors, area etc. - or a more basic solution should be considered?
1.2 Should the system be configurable enough and leverage a Rules Engine to implement - partially or not - the Mathematical model and Heuristics; or should it simply hard code them?

2. Considering that it should scale to millions or billions of elements
2.1 How the different computations should be parallelized i.e. more or less design a parallel version of the sequential steps using well known techniques such as pipelining, partitioning etc. At least the specify the principles and approach
2.2 What parallel/distributed computing platform should be used i.e. Grid, cluster...
2.3 What supporting technologies should be leverage/used i.e. messaging, Agents...

3. What existing components/technologies should be leveraged - when possible - or should in house development be prefered?


The answer to 1 depends - besides the CS/IT culture of the designers - on the budget, resources, the goals and Use Cases. Unless the goal is to create a reusable, configurable and extensible product - and I don't think it's the case here, because it's an application - the simplicity of the solution will probably be preferred and determine the answer i.e. no fancy data repository or Rules Engine.
In both cases, it is of no particular challenge from a CS/IT perspective as they are well known and understood activities.

The answer to 3 mainly depends on culture and my answer is "always??:" leverage existing components - Open Source or commercial.
Again, I don't see any challenge as
1. Product/tool/component evaluation has been simplified because of compliance to standards and, in each tool space or for each standard, their exist well known implementations. Also, their exists readily available evaluations either by analysts - i.e. The Gartner Group, Forrester Research etc. - or other specialists
2. The architecture of the system is quite simple in this case

The answer to 2 is mainly technical in nature, and if the answer to 3 is to leverage existing components, it would not be considered a CS/IT challenge.


Implementation Phase
Considering the simplicity of the system, the implementation/coding should be relatively straightforward and would not be considered a challenge, from a CS/IT perspective, when it leverages already existing components and libraries, Similar, computationally intensive but architecturally simple, parallel computing systems have been built for many decades; however, in the past, the necessary supporting technologies/components - or the "plumbing" - was not readily available and had to be developed in house.
The main difficulty, for the average developer, would be to troubleshoot a concurrent system.

Did I miss anything that would be a CS/IT challenge?

[karoaper]

Re: S.o.s!!

Sorry for the delay in my response.
I think there's a bit of a confusion about what is meant by an algorithm. I meant it not in terms of a software construction or program code, but in terms of providing solutions to complex (mostly) mathematical problems.

Certainly software construction has its own set of challenges, which I think are more in the domain of engineering, such as scalability, testing, cost, etc. The algorithms I was talking about deal with such challenges as reducing memory and time complexity, reducing computational errors due to scale issues, etc.

[Sip]

Re: S.o.s!!

While we are on the topic of "non-challenges" from the computer science perspective, I think this article is definitely worth a look: Quantum chip rides on superconducting bus. I just can't help but wonder how much of an algorithmic non-challenge it's going to be to go from thinking in terms of "1 or 0" to also be thinking in terms of "1 and 0" when dealing with these qubits.