These are chat archives for non/algebra

9th
Jan 2016
Anton Kulaga
@antonkulaga
Jan 09 2016 00:45 UTC
What I really lack in scala right now is a collection of good ODE solvers (esp. for stiff problems) or maybe binding to Fortran/C++ packages like lsode. And computational biology is often about ODEs =(
The other problem with Scala adoption is that programming in research world is very different. Most of the work is done by one to several scientists working on the same script in matlab/python/R and it is even hard to explain them why they need classes and why calling all variables with greek names (all those kappa, gamma, tetta) is not always a good idea. In fact sometimes it is good if those letters are used in the paper
And in most of the cases they actually do not need classes and other stuff much. What they need is a compact form to represent vector/matrix computations and at the moment matlab/octave/julia are better then breeze in this regards (less verbose)
Pascal Voitot
@mandubian
Jan 09 2016 10:47 UTC
@antonkulaga actually, while working in a company with datascientists, I've (re)discovered how much both development worlds (Research vs Prod) were far away and I think this is a shame because the gap shouldn't so big... But as soon as I've shown to a mathematician the way I was developing with algebraic concepts & types to prove a lot more in my code, they've seen the interest of it with respect to their work, it's clear. But, lately, implementations of pure maths algo in production with realtime/distributed computations, critical data & 24/24 availability become more frequent than before... so we might have a luck to get closer from each other even if it might mean hiding terrible C code in a lib behing a FP oriented API
Pablo Pita
@pleira
Jan 09 2016 12:06 UTC
@antonkulaga I am experimenting in my personal scala project with using variable names close to those given in the formulas. That is, the code should look similar to the actual algebraic expressions. I have literal greek letters and also exponents embedded in the variable names. Here is a method to have a look. I find the formulas look quite readable in the code and it is worth the effort with "cut&paste" the variables often as you can not easily type them from the keyboard.
I agree with @mandubian that development world and research are far away. I believe that as both worlds have so many complex issues by themselves that it is normal so. When implementing the algorithms, both worlds come together and indeed, there is a good chance that more scala libraries would be developed to provide new functionality.
Denis Rosset
@denisrosset
Jan 09 2016 12:14 UTC
@antonkulaga it's best to have greek letters in your code when they refer to a published article; the article then serves as the best documentation possible