Madagascar development blog - Comments

Esteban : Program of the month: sfderiv

nospam@example.com (Esteban ) — Sat, 18 May 2013 13:42:15 -0700

I just added a new example (book/rsf/rsf/sfnderiv) comparing sfderiv with sfnderiv, a new
program based on Fornberg, 1988 paper.

I demonstrate that both programs have equal impulse response for first derivatives.

For higher order derivatives sfnderiv is better that piping several passes of sfderiv.

Karl Schleicher: Events of the year

nospam@example.com (Karl Schleicher) — Fri, 03 May 2013 11:22:44 -0700

I will contribute some open data and work on entries to the migration gallery. My contribution might be new row in the matrix or just filling in one of the odd missing spaces. I will attend the SEAM workshop in May and hope to get some of the data in time for the workshop. If that data is not distributed by that date, I will work on some other open datasets.

Esteban : Program of the month: sfderiv

nospam@example.com (Esteban ) — Wed, 01 May 2013 14:21:40 -0700

I have been looking into this paper for a class in numerical solutions to PDEs to get optimized coefficients.

This paper gives good approximation to first order differentiators. What about second to n order derivatives?
To get around this I use Fornberg, 1988 paper which gives similar results for first order derivatives, and optimized ones for higher order ones (which is much better than doing d2fdt2.rsf).

After I am done with classes, I will put something under my user.

Cheers,
Esteban

Construction Management: Wavefront construction

nospam@example.com (Construction Management) — Wed, 27 Mar 2013 22:03:56 -0700

I like your post and it really gives an outstanding idea that is very helpful for all the people on the wavefront construction. Thanks for sharing.

Nick Vlad: School on Reproducible Science

nospam@example.com (Nick Vlad) — Tue, 19 Feb 2013 18:32:36 -0800

Til lykke!

Nick: FFT sizes

nospam@example.com (Nick) — Sat, 26 Jan 2013 11:16:53 -0800

A wrapper to the optimal-size finder ( sfpad2nextfastsize ) was added to the development version, so this functionality can also be accessed from Python and shell scripts

Nick Vlad: The necessity of object orientation in geophysical software applications

nospam@example.com (Nick Vlad) — Fri, 21 Sep 2012 19:23:02 -0700

Sorry for the confusing term. By it I mean classes and methods that are short and simple, with no complex relationships and inheritances, just one step up from classic procedural programming, so they are easy to understand, write and debug even by a novice.

Filippo Broggini: The necessity of object orientation in geophysical software applications

nospam@example.com (Filippo Broggini) — Tue, 11 Sep 2012 14:54:28 -0700

Dear Nick,

I am not sure if I properly understand what "light" object orientation is. It is the term "light" that confuses me.

Thanks a lot.

Bill Symes: The necessity of object orientation in geophysical software applications

nospam@example.com (Bill Symes) — Wed, 22 Aug 2012 16:29:23 -0700

IWAVE is built in this "light OO" fashion, plus a bit more. The backbone consists of structs realizing key FD types and associated functions for manipulating them - lightweight "objects", with static members acting as private data, as you describe. IWAVE also uses a form of runtime polymorphism (inheritance): for the central structs, the corresponding functions are accessed via function pointers which are members of the struct, and must be initialized by calling a "constructor", so that these "class methods" are accessible. This setup realizes an abstract type with concrete subtypes. IWAVE uses it to abstract a large part of the essential parallel FD code so that it can be shared by all apps of the class.

I also believe that there is a role for genuine, gold-plated OO in contexts where great differences in abstraction level exist. The main opportunity in our business is simulation-driven optimization, also known as inversion.

Bill

Sergey Fomel: tkvpconvert

nospam@example.com (Sergey Fomel) — Fri, 20 Jan 2012 08:49:00 -0800

Not at the moment but we could add it if needed.

Usage: vpconvert [format=] file.vpl [file2.vpl file3.vpl | file.other]

Toastar(Mike Stone): tkvpconvert

nospam@example.com (Toastar(Mike Stone)) — Fri, 20 Jan 2012 07:47:33 -0800

Dumb Question...

Does vpconvert support stdin/stdout?

Sergey: Science Code Manifesto

nospam@example.com (Sergey) — Wed, 19 Oct 2011 17:54:23 -0700

I can understand the reluctance of some authors to release code and/or data. However, your points 1 and 5 look invalid from the scientific-method point of view.

According to the current practice, you cannot publish a new mathematical equation or theorem without providing a derivation or a proof. Arguments like "my derivation is not cleaned-up yet" or "I don't my competitors to see it" will not be accepted by journal reviewers or readers. If we want our computational research to be scientific, the same standards should be applied to publishing computational results.

Mike Stone (Toastar): Science Code Manifesto

nospam@example.com (Mike Stone (Toastar)) — Tue, 18 Oct 2011 11:03:10 -0700

Well my personal viewpoints are fairly copyleft, I want to play devil's advocate here for a second and highlight some points that were made on the AGU Linked-in Forum

1. The author might be reluctant to share because their code might be workable but needs clean-up and documenting to be readable.

2. Copyright: Most journals require you to turn over copyright to them, With the text of an article this isn't as big of a deal as it is with working code.

3. Including Code adds a burden to the peer review process.

4. Data, What if the dataset used was too large to reasonably distribute? What about proprietary data? your much more likely to get a corporation to approved 2-3 screen-shots then an entire cube.

5. Competition, if a journal requires the code to be released it might discourage commercial entities from publishing as it would allow competitors a chance to catch up faster.

----------

Here's my pie in the sky Idea, What's really needed is a new open access journal. The copyright would be fluid, With the minimum of the owner granting a enduring use license; but strongly suggesting a creative commons license.
Papers would be ranked in to several categories:

1. Unreviewed papers, Authors who didn't pass the peer review process could resubmit to this section.
2. Peer reviewed section without code or data.
3. Peer reviewed with data and code available to the reviewer
4. Peer reviewed with Code and Data available.

This structure would allow you to sort the academic papers in class 4 from the sales pitches in class 2 and 3. Also Class 1 would let you get the occasional gems without muzzling the crackpots, but at least give you a way to filter them out.

archit gupta: An idea for Google Summer of Code 2012

nospam@example.com (archit gupta) — Tue, 04 Oct 2011 20:20:26 -0700

can anybody give the information that how can we become the part of GSOC 2012, i am keenly interested in it

Nick Vlad: Which country has the most Madagascar users?

nospam@example.com (Nick Vlad) — Sat, 17 Sep 2011 14:24:37 -0700

Analytics is a fun topic.

It would be most meaningful to normalize by the number of geophysicists in each country, but that is hard to come by. SEG membership is a proxy, but not in all countries that is widespread, or in the same proportion with the total number of professionals. SEGE+EAGE membership might be at first sight better, but some people hold double membership... Not easy :-)