User login

You are here

Sharing data by theory/computation mechanicians

Pradeep Sharma's picture

I noted with interest the recent iMechanica blog entry regarding the fact that only a small minority of experimentalists are willing to share raw data. I was somewhat surprised. I instinctively felt that theoreticians and comptutational scientists would have a different attitude. However, just a couple of days ago, a friend alerted me to an article in the Physics Today (--the mouthpiece of the American Physical Society) regarding an ongoing feud related to a peculiar co-existence of two phases in water. At its heart, the problem being described in the Physics Today article is a mechanics problem and makes for a gripping story. Also, it underscores the point that there has to be greater transparency in computation. At least in regards to the use of atomistic potentials (where great dangers lurk), the openKIM project started by Ellad Tadmor and Ryan Elliott provides a resolution.

Comments

Douglas P Holmes's picture

I would certainly welcome greater transparency across all aspects of mechanics research - experimental data, computational code, and mathematical calculations (e.g. Mathematica notebooks). I think the Jupyter Notebooks (laregly use with Python) are a great step forward on the computational side. For experimental data, what do we do with terabytes of high resolution photographs that are later analyzed with image processing tools (tools that always aren't shared nearly enough). The DataVerse project seems intriguing, but I'm just learning about it now (https://dataverse.org/).

The challenge is always the same - sharing code/data takes time and energy to do well, and there is no strong incentive. I, for one, would gladly welcome a change to this status quo.

Pradeep Sharma's picture

Hi Douglas, I agree. Thanks also for pointing out DataVerse and Jupyter Notebooks. I will look into both. This has prompted me to reexamine how I do business and see what I can do on my side. The key issue is your observation that doing it well takes time and effort and there is not much in the way of incentive other than a personal sense of communal service.

Dear Douglas,

Across. the. board. ? ... What precisely do you mean by that?

Do you---by any chance---mean to say: mixing up the private with the public, as in reference to sheer size of the computational data and purported scholarly. interests. ? [The reference is to what I said here [^]]

[Indian languages]  ``Tch!''

--Ajit

 

 

Emilio Martínez Pañeda's picture

There is another benefit, besides enabling reproducibility, intrinsic to code sharing in computational mechanics. Essentially, by sharing our codes, we are providing the community with tools that should speed up scientific progress. Based on these two motives (allow people to reproduce my results and, hopefully, help them in their research efforts), I decided some time ago to release the codes employed in all my papers (www.empaneda.com/codes). While this is a time-consuming process (I had to clean the codes and write a documentation), it has proven to be rewarding; I received many e-mails from users around the globe thanking my openness and explaining what they have achieved by using or extending my codes. 

​​​​Emilio Martínez Pañeda​

Self-financed or publicly funded?

--Ajit

Emilio Martínez Pañeda's picture

Hi Ajit,

My research endeavours are primarily financed by public funding agencies but I have also conducted research for private companies. I can see the struggle of sharing all the tools in the context of an R&D contract with a company; in my case, I typically develop the tools as part of a publicly funded fundamental project and these tools (that have been shared) are then used in the more applied industry-backed projects (frequently, by a student that could end up working for the company).

Emilio Martínez Pañeda

Pradeep Sharma's picture

Hi Emilio, this is an important point and I laud you for your own generosity. After some initial stage, once the key papers are out, I believe it to be a healthy practice to make the codes available publically. My group is in the process of doing the same with some of our own codes---once they are cleaned up. I see this happening more or more in the computational materials science community and several of our mechanicians have done this in the past. A good recent example is the finite element based DFT code released by one of our own mechanicians: Vikram Gavini.

Subscribe to Comments for "Sharing data by theory/computation mechanicians"

Recent comments

More comments

Syndicate

Subscribe to Syndicate