Scientific Computing Category

Scientific computing is a large group of users who do applied mathematics, numerical linear algebra, numerical differential equations, mathematical modeling, etc. One of the first questions on the discourse site was on FEM discretization of PDEs. With how tailor made Julia is to scientific computing, it should probably have its own domain.

While some may not like the name, it’s always been named either this or Computational Science, see anything you count as a reference. The StackOverflow for this group is named Computational Science but is prefixed with “scicomp” for scientific computing. It’s pretty much agreed on what “Scientific Computing” means.

I feel that Scientific Computing is too broad to be effectively useful as a category. To quote the wikipedia entry:

Computational science (also scientific computing or scientific computation (SC)) is a rapidly growing multidisciplinary field that uses advanced computing capabilities to understand and solve complex problems. Computational science fuses three distinct elements:

  • Algorithms (numerical and non-numerical) and modeling and simulation software developed to solve science (e.g., biological, physical, and social), engineering, and humanities problems
  • Computer and information science that develops and optimizes the advanced system hardware, software, networking, and data management components needed to solve computationally demanding problems
  • The computing infrastructure that supports both the science and engineering problem solving and the developmental computer and information science

In essence most of the debates in the Domains category is going to be about scientific computing.
For me categories should be as broad as possible, while being as specific as possible. As you said

I am not categorically opposed to create a category for Scientific Computing, but I would prefer that the different sub-topics of Scientific Computing, get their own categories, e.g. Bio, Physics, Numerics, .

I would like to suggest Numerical Methods for this. Wikipedia equates this to Numerical Analysis but those of us with some mathematical training probably think analysis means the theorems that justify the methods. The Wikipedia article says NM/NA includes

  • Computing values of functions
  • Interpolation, extrapolation, and regression
  • Solving equations and systems of equations
  • Solving eigenvalue or singular value problems
  • Optimization
  • Evaluating integrals
  • Differential equations

It seems that optimization is the only one of these with enough activity to deserve its own category here (for now), but I second Chris’ point that the many posts in this general area deserve a domain.

1 Like

Those aren’t subtopics of Scientific Computing as it’s commonly used. Computational Biology tends to be more Bioinformatics / Data Science, though Computational Physics tends to be highly related to Scientific Computing, and Numerical Analysis is addressed below.

The field numerical analysis is the mathematical part and scientific computing is the application, i.e. putting it on a computer and the algorithmic issues associated with it. That’s at least how journals tend to distinguish it. A better source would be SIAM:

https://www.siam.org/journals/sinum.php

https://www.siam.org/journals/sisc/redefined.php

Maybe it’s the best tern we have that those outside the field won’t misunderstand (though I would’ve thought scientific computing is well-established enough to be a specific term. The real problem is it’s relatively common for people in Data Science don’t tend to know about Scientific Computing, I’ve had this discussion a few times and know it’s a confusing term to those who aren’t in the discipline).

A big number of broad categories and less broad subcategories down to narrow ones of low traffic should not be a problem, as long as people reading their titles can easily distinguish them. There will always be threads that are relevant to more than one categories, that’s not a problem either. The problem is when subcategories are relevant to more than one categories. I believe that’s the meaning of “specificity” @vchuravy suggests.

There are more than one ways to construct a tree of categories. In multi-word category names, the important word to construct a tree after is the first one. For example:

  1. Scientific Computing” could house “Biology”.
  2. Computational Science” could house “Computational Physics”.
  3. Numerical Analysis/Methods” could house “Differential equations”.

Mixing competitive ways would confuse users, so it’s important the name choices to be consistent across the tree. Even if an individual name is not the most fashionable in the respective scientific community, people will still understand it when they read it. Therefore, instead of discussing individual category names in separate threads, we should use a wiki to discuss all names in parallel. Unlike with package names, where each team decided a name individually, the effort here in Discourse is to integrate the ecosystem, so as to make its value higher than the sum of its parts, for the benefit of the whole Julia community.

No it can’t, not without changing the meaning of scientific computing from its common use to its literal words.

No, just the talk about methodology for computing specific equations. That’s where the overlap is.

It can, but in the way it’s actually used it’s not about implementation of differential equations. I am pretty sure more people who are here are talking about implementation and using packages, not convergence proofs, which is why I’d call it scientific computing.

1 Like

My primary concern is to minimise confusion for new users. How about we add two domains, one for Numerics and a second one for Modelling?

The first one is to talk about numerical algorithms, and the development thereof. And the second one is application focused, on how to apply computational methods to a certain problem and which tools to use.

I could see that being sensical.

1 Like

I created Numerics and Modelling&Simulation, the description posts are wikis so please feel free to edit them.

I suspect we may eventually want to split modeling and simulation into two categories, but that may be a ways off.

This is my final post on this thread, as my argument is not about the suggested category. I’m interested not in people’s preferences, but in the big picture of things.

It definitely can, it’s only a matter of a few clicks. But these are not suggestions for names, they are mere examples of the proper relationship between the name of a category and the names of its subcategories. In such cases there is no such thing as a perfect name, cause they all have their disadvantages, cause human language itself is imperfect, as are the people using it.

You focused on the wrong part of my post, namely the examples. People could endlessly argue about proper calling, especially when they don’t make an effort to consider alternative points of view. And majorities come and go, their advice rarely passes the test of time.

My point is for everyone to raise their heads a little and try to look beyond the limits of a single category or package or organization they happen to spend most of their time in, so as to cooperate in finding names that in the long term will benefit the whole community. This isn’t the case so far. If people refuse to look around them, fragmentation can easily take place even inside a single forum.