Forums

Complexity of DSP magazine articles - Opinions requested

Started by Rick Lyons 6 years ago9 replieslatest reply 6 years ago208 views

The IEEE Signal Processing Magazine has a monthly column titled: "Lecture Notes". Those Lecture Notes articles are intended to be "easily accessible, being proper for our DSP students and young professionals."

The magazine's January 2017 issue contains a 'Lecture Notes' article titled: "Compressive Privacy: From Information/Estimation Theory to Machine Learning".  Counting algebra expressions such as 'i < L' and

equation_91385.jpg

as equations, that Lecture Note contains

* 57 standalone equations,

* approximately 72 equations embedded in the text, and

* approximately 10 equations embedded in the figure captions

for a Total of roughly 140 mathematical equations in one Lecture Note article!!

It seems to me that 140 equations in one magazine article are too many equations for a reader to "keep track of."  What's your opinion?

[ - ]
Reply by arangurenjJanuary 26, 2017

No doubt, it is overwhelming much.

One of the reasons I dropped my subscription (I was paying it myself), because not really much was of practical use.

OK, the title is "lecture notes". Probably targeting people more interested in the academic aspect of DSP, rather than in the "mundane" practical implementation for solving real-life problems.

Just an opinion.

[ - ]
Reply by MichaelRWJanuary 26, 2017

Good morning,

I think this is a somewhat common ailment of most technical articles today.  Non-technical articles too.  I get the sense that the authors have the need to treat every aspect of their topic to ensure that the article has enough "hooks" in order to attract readers and, perhaps, highlight their intellectual "chops" or prowess.

Sometimes these articles are useful starting points for people new to a field, a guide to the terminology and ideas.  In these cases the authors have been careful to adhere to this goal and have written the article appropriately, expressing the general ideas of the area and including only relevant references.  They also keep jargon and the number of equations to a minimum.

At times I still find myself trying to read such articles, but I'm getting better at understanding the fact that if the article is about the "100 Important Details of Feature Selection for Optimal Machine Learning", for example, that there are better, more succinct resources available.


Michael.


[ - ]
Reply by drmikeJanuary 26, 2017

I love math, and the higher the equations to text ratio the better.  But that is definitely because I intend to become an expert in whatever it is I'm learning about.  From the description "proper for our DSP students and young professionals" I would agree it fails, it is aimed at experts, or people with broad experience who want to become experts.

There are times when math is the most succinct language.  It should only take a few equations to describe a problem.  Solutions take a lot more, but an introductory article isn't going to have space to really deal with that (unless it's 20 pages long). 

I'm betting a lot of those equations are redundant.  This is a style issue in a way - do you want to refer to an equation 3 paragraphs back or just repeat it so the reading flow is not interrupted?  Lots of times the same equation is written with different variables on the left - again redundant.  If it's 140 independent equations, nobody can keep track of that, let alone students!

Does the math flow with the text in a way that makes sense?  Is it written so you don't have to juggle 5 equations in your head at once to see how a new equation is derived?  If it flows, then maybe it isn't too much.  But if it is just a lot equations which are covering a lot of ground so the article fits in as few pages as possible, it's not for students.

Dr. mike

[ - ]
Reply by kazJanuary 26, 2017

the concepts must first be explained without any numbers then fortified/quantified with equations. Not the other way around.

Kaz

[ - ]
Reply by Rick LyonsJanuary 26, 2017

Hi Mike.  The article I referred to was 10.5 pages in length.

[ - ]
Reply by drmikeJanuary 26, 2017

14 equations per page is too much.  I suspect I would have slept through that lecture!  All the technical papers I have on my desk have 5 to 7 per page, and these are not for students.  As Fred points out, "We publish to impress, no?" - sometimes impressions are negative.

[ - ]
Reply by asnJanuary 26, 2017

A few years ago, I tried out the IEEE signal processing magazine subscription for one year, but didn't like the writing style. The articles read like journal publications which were (in my opinion) almost impossible to follow without a solid background in the area. Many authors jumped directly into the advanced concept, and didn't really provide any practical application info.

T The UK variant (IEE computing and control) was significantly better, by providing well structured articles which were also technical, but I always felt that there was take home message. Unfortunately, the IEE turned into the IET (in order to attract a wider audience), and as such the quality of the articles went downhill. 

Rick, perhaps you can keep up the good work?

-Sanjeev.

[ - ]
Reply by Fred MarshallJanuary 26, 2017

Like Kaz, I believe that the mathematics should follow physics / understanding.  It's a bit like sunglasses: it improves the image only if the image is there to begin with.  

I would admit that there could be situations where pure mathematical expressions can lend insights to some.  I believe those occasions are rare compared with the opposite.  

I have the notion that the simpler the equation, the more likely it might be generalized in a useful way.  E=mc^2 comes to mind.  

I can well imagine making mathematical mistakes or creating overly simplified mathematical models due to a lack of understanding.  But then, much of my universe is made up of things that have to be understood.

So, pages full of complex equations that are made to be as terse as possible are usually of no interest to me.  Page charges may be one reason for things to be terse.  Elegance of notation is another reason for things to be terse.  Maybe we should adopt the phrase: "Everything should be less terse as possible and no less terse"  But some stuff can be read if you understand the notation.  I find myself speaking in my head:  "C is contained in E such that ..... etc."  Then there are the cases where there are pages and pages of matrices.  Not terse and not readable really either.  For mathematicians some things might be important when constructing a proof.  For practitioners of the mathematical sciences it's often nonsensical fluff.  Oh!  But it is impressive!!  We publish to impress, no?  

Oh?  No?  :-)


[ - ]
Reply by sbaby171January 26, 2017

Yeah, I agree with you Rick. To illustrate their point to students and young professionals, a magazine article meant should have a high word to equation ratio.

Of course this is simply my opinion, but I believe much of the writing should be focused on describing the main concepts through the use of analogies and metaphors rather than equations. The easy way to write is simply plaster equations across your paper, I find that writing technical engaging papers calls for more skillful writing. The writing should be creative as well as technically sound if you are writing to engage new young audience.

Lastly, the paper should be the catalyst that causes them to pick up a textbook. It is in the text book where the student will become more familiar with the material. 

The equations may just deter the semi-interested student or young professional. Thus, by using analogies and metaphors, it offers the opportunity that, should the student not going any further than the magazine article, they would walk away with a working super-level understanding of the topic at hand.