In the December 2006 edition of IEEE Computer, Simone Santini, from the Universidad Autónoma de Madrid asks “Standards: What Are They Good For?” The gist of his argument is that using concepts like XML in Computer Science is harmful. He argues that there is no such thing as XML technologies since XML, like all standards, is nothing but a rather slimy mix of politics and industrial concerns. He says that no W3C standard is deserving of a Computer Science paper. (Does he know that W3C never issued a standard?)

Why? Mostly because, “having academia operate on industrial principles makes about as much sense as having the industry operate on academic ones.” Interestingly, he argues that if Computer Science had been as focused on standards 40 years ago as it is now, we would still be programming in Fortan and Cobol on OS/360 operating systems.

You see the problem in his argument? First, academia should not operate on industrial principles, but yet, somehow, academic research is meant to lead to industrial progress (such as Java, C++ and Windows XP?).

I like the guy, for sure, just because he dares to publish this in IEEE Computer… an engineering publication. Do you know people more obsessed by standards than engineers? He is not going to win a popularity contest and I am amazed he managed to get this opinion piece to press.

Here are some arguments I would like to submit to him:

• There is zero evidence that having researchers interested by standards, such as people studying XML, is having a harmful effect on the pace of technology. It may harm some theoretical research work, but I seriously doubt it.
• Computer Science, as a branch of Mathematics, should not care about XML or the W3C. But academic research does not stop at Computer Science. Information Technology is an increasingly important discipline and it cares very much about Web standards. Moreover, Computer Science is sometimes considered and engineering discipline (think about “software engineering”) and engineers need to care about standards . Finally, Computer Science should be an empirically discipline and not just a branch of Mathematics, and in this respect, it should care about standards and study them from a scientific point of view (maybe to improve them!). For example, XML is a bit more than just an “unranked labeled tree”: it is one of the most interesting phenomenon in Information Technology that I know of. Choosing not to study XML would be like choosing not to study the Web.
• Not all standards are ugly compromises. It often goes down that way, but some standards are fun and interesting.

So, my answer is clear: yes, academia should care about standards, despite Simone Santini’s point of view.

(Disclaimer: I do not think that any of my papers have been standards-centered, and most, if not all, are not standards-aware. I do not generally write papers about XML or XQuery or XHTML. My point though is that such work is perfectly legitimate.)

[photopress:lohan_avion.JPG,full,pp_image][photopress:louka_police.jpg,full,pp_image]

I think this is our first snow man picture!

[photopress:PC2600191.JPG,full,pp_image]

I will present my paper A Better Alternative to Piecewise Linear Time Series Segmentation at SIAM Data Mining 2007 in April. The paper is available on arxiv.org (pdf).

Here’s the abstract:

Time series are unstructured data; they are difficult to monitor, summarize and predict. Segmentation organizes time series into few intervals having uniform characteristics (flatness, linearity, modality, monotonicity and so on). For scalability, we require fast linear time algorithms. The popular piecewise linear model can determine where the data goes up or down and at what rate. Unfortunately, when the data does not follow a linear model, the computation of the local slope creates overfitting. We argue against declaring as flat, intervals where the slope is not significant. We propose an adaptive time series model where the polynomial degree of each interval vary (constant, linear and so on). Given a number of regressors, the cost of each interval is its polynomial degree: constant intervals cost 1 regressor, linear intervals cost 2 regressors, and so on. Our goal is to minimize the Euclidean (l_2) error for a given model complexity. Experimentally, we investigate the model where intervals can be either constant or linear. Over synthetic random walks, historical stock market prices, and electrocardiograms, the adaptive model provides a more accurate segmentation than the piecewise linear model without increasing the cross-validation error or the running time, while providing a richer vocabulary to applications. Implementation issues, such as numerical stability and real-world performance, are discussed.

Mindset is a research program to train software to recognize commercial pages. One application of this tool is that you can try to exclude commercial pages out of the result set.

Of course, if this is as good as spam filtering, people will only be partly happy with the results. And yes, there are many commercial Web sites trying to pass out as non-commercial. Everyone is out to sell something, afterall.

Interesting question: would you ever want to do the reverse, that is, exclude non-commercial content?

(Source: Turney.)

Roger Schank, a math wiz, says we do not need to teach math and science.

What (…) makes no sense is the idea that math and science are important subjects. You can live a happy life without ever having taken a physics course or knowing what a logarithm is.

On the other hand, being able to reason on the basis of evidence actually is important. Thinking rationally and logically is important. Knowing how to function in a world that includes new technology and all kinds of health issues is important. Knowing how things work and being able to fix them and perhaps design them is important.

Lets get serious. We don’t need more math and science. We need more people who can think.

Of course, while I agree with his point, he is being a bit too hasty. If you want to run a business, you need to know that if you save 10% and then save 10% again, you do not save 20%. You need to know that if you sell something US$10 and US$1 = CAN$1.5, then you sell it US$15. So, we do need to teach mathematics if only because everyone (at least in Canada) is responsible for filling out tax forms. Yes, you can get your tax forms filled out by an accountant, but, in principle, you are the one responsible for any mistake made.

You do not need to know what a logarithm is? Depends what you do for a living. If you are a programmer and you need to sort a bunch of entries, you need to understand that the first algorithm you will think up (typically Bubble sort) is not going to cut it if you have to sort 10,000,000 entries. If you want to really understand the issue you need the concept of logarithm.

What about trigonometry? Most people working in a factory doing non trivial work need a basic understanding of trigonometry.

I must admit that I have little use for the chemistry I learned, but then, I have little use for the geography either. Who needs to know where is Val d’Or?

Where he is right however is that math and science are not as important as some lobbies make it out to be. I know a lot about how to solve nonlinear differential equations. Way more than I need considering I haven’t seen a nonlinear differential equation in nearly 10 years.

But to be fair, education takes a long time to adjust. My education was modeled after the space race. We were all to be rocket designers or astronauts (or maybe cosmonauts if you were a pessimist). So Physics, differential equations, algebra, and so on, were thought to be central. It turns out that it is a fringe subject. Few people work for the space industry.

In Computer Science, for a long time, we thought that we were limited by our computing resources so designing extra efficients algorithms was very important. As it turns out, Tim Berners-Lee convinced me that we are not building the future out of algorithms.

So, one generation teaches the next what they think is most important. The old generation is almost always wrong. Yet, it almost seems not to matter because apparently, we manage to go forward.

Nothing to worry about here. The Americans will not go bankrupt because people in Bulgaria are twice as good at math. If you learn math, you probably learn to think straight, but there are other ways to learn to think straight.

(Source: Downes.)