What answers should be "Answer of the week"?

In MATLAB Answers, many answers are good enough to satisfy the questioner; but some go further. They may take extra care in explaining some MATLAB technicality or the reason their solution works. Or they may go beyond the question that was asked to address a broader issue. Such answers are all the more remarkable because the contributor knows they are in a race: someone else may post an adequate answer and have it accepted before they are finished.

That came home to me when I saw this answer by John D'Errico to the question find all answers to the equation. The problem was how to find the roots of exp(-0.2*x)=(x-2)*(x-3)*x. The accepted answer was pretty good, using a sensible approach to find three real roots. But D'Errico went beyond that, posing the question "How will you KNOW that you have found all solutions?" and demonstrating a root deflation method that returned several complex roots in addition to the three real ones. But how many people are ever likely to see this answer?

The File Exchange has the Pick of the Week blog to highlight especially good contributions, but MATLAB Answers doesn't have anything equivalent. That is why I have created this question. What answers do you think deserve highlighting? The answer can be from as far back as you like, and you don't have to wait a week after the previous post. Just share it with us so other people can benefit from it!

Answer 1

[John D'Errico]

This seems like a nice idea. It might encourage people to spend just a bit more time writing an answer. That can never hurt, and it can only help the person who needs the answer to their own question.

I see two viable solutions.

1. Someone at TMW, probably on their Matlab Central team, would post a blog entry, once a week. This would entail them reading through many Answers posts every week. Yes, they could invite e-mails nominating solutions to highlight. But it would take some effort.
2. Use this very question. Every week, or whenever you feel moved by an excellent answer, post an answer to this question, that highlights that answer (with a link). Add enough context about the question and what you feel to be the best features about the answer.

I vote for option 2. It is trivial to implement, whereas option 1 would require someone from TMW to be willing to dedicate their time and energy. And since anyone can add an answer, you effectively have hundreds of eyes available for the search, so no single person needs to invest a large amount of time.

[Andrew Newell]

I hope no one thinks that they need to wait a week! I posted this late at night on an impulse, and with a little more time to think I have reworded and clarified the question. My vote is option 2 for now, with no prejudice against option 1 being adopted at a later date. But there is something fitting about contributors thanking each other for their work.

Answer 2

[Andrew Newell]

Here is another one from a few years back: this answer by David Young to the question, What is the fastest way to do 2-D convolution in Matlab? Young compares three functions: the MATLAB builtin conv2 and two contributions in the File Exchange: convolve2 and conv_fft2 He creates tables of time against grid size for various kinds of problems.

[David Young]

Thanks Andrew. I suspect my answer is now rather out of date - I have been meaning to extend and rerun the tests - now maybe I'll get round to doing it!

Answer 3

[Andrew Newell]

Here is John D'Errico again, answering the question: Fibonacci program makes matlab go busy forever. The user wonders why a recursive function to write Fibonacci numbers gets slow so quickly, and gets a dissertation in response. Don't miss the comments to this answer, where he describes a surprising relationship between Fibonacci numbers, matrix determinants and Chebyshev polynomials.

Answer 4

[Star Strider]

What criteria do you want to use? Brilliant but verbose essays that appeal only to the Editors and other cognoscenti? Short, succinct responses that reduce a complicated Question to a simple (and in retrospect, obvious) Answer? A mathematical ‘proof’ or derivation that educates the ‘Querent’ in the broader implications of the Question? A condescending (albeit correct) explanation?

Other than a direct answer to a homework question (to be eschewed if at all possible because it’s rather blatantly cheating), would not the ²best’ answer be the one that educates the Querent to the greatest degree and has the greatest value to other potential Querents?

I, honestly, have no firm idea. I would propose that ‘Best’ Answer or the ‘Answer of the Week’ be the Answer that most likely stands the test of time (absent MATLAB ‘upgrades’ and ‘improvements’ that might make them obsolete in a future release) but that educate in the broader sense the underlying mathematics involved in mathematical programming, and specifically, of course, MATLAB programming.

It is easy to answer a specific question, but does that answer do anything more? If not, it is not a candidate. Is the answer one that stimulates the Querent to do other research on his or her own? If so, that may be the winner. That my criteria are the primarily ‘pragmatic’ or secondarily ‘utilitarian’ approaches is not coincidental. They are my weltanschauung.

So what is the ‘Answer of the Week’? In my pragmatic view, the Answer that stimulates the most subsequent exploration and understanding by the Querent and all who read the Question and Answer. We can, as the AAMPAS does, pat ourselves on the back and tell us how wonderful we are. But I would like to leave that decision to the Querents and other end-users to determine. This not only puts the Querents first, but tells us what they want to know, and how they want to know it. (I realise that this includes those who need to produce a Nobel-quality paper by tomorrow morning, but it also includes brilliant people who quite literally ‘never thought of that approach’ and learned something — possibly very valuable — from it.

[Andrew Newell]

Star Strider, I agree that I have been favoring 'brilliant but verbose essays' and that more concise responses may sometimes be more useful to the 'Querent'. It's partly because assessing this kind of essay is easier than judging the usefulness of answers to Querents.

When I posted this Question, I was hoping that other people, especially Querents, would nominate answers, but so far I'm alone here. The only criterion for including an answer on this list is that someone took the time to nominate it.

[John D'Errico]

In my case, I admit that I tend to go overboard, trying to answer all possible followup questions that should pop up. Us (find him on the FEX) used to call my FEX codes over-complete. But this also allows me to not need to worry so much about needing to come back and be forced to make multiple followup responses. I think it is just a question of style myself.