Hi everyone :) I am looking for a bit of advice regarding the implementation of parallel tempering in C programming. I have absolutely zero experience with C but I have been tasked with writing an algorithm despite this, of course I am eager to learn and I really appreciate any help! I have asked a couple of questions but I think they violate the "off-topic or too vague" guidelines for the site, likely because I am really new to all this!

Okay, So my data structures will hold an awful lot of variables, think about bond lengths (floats) bond angles (floats), number of atoms (int), names of molecules (characters) and then multiply them by maybe a couple of hundred atoms or so (there are lots more variables than that, it's just an example).

Unfortunately the department is adamant that it be C to fit in with existing algorithms for global optimisation techniques

otherwise I would have to re-write entire program which is something I am really not capable of doing. I know, behind the scenes we chemists have no clue at computing haha

fair enough ;)

oh indeed, although I don't work with any nasty chemicals haha, theoretical chemistry all the way!

really :/

One of my colleagues has an undergrad masters CS student working for him, so I have been chatting to him about C trying to learn as much as I can

oh dear that sounds like fun

hmm, well I'll detail my problem and then detail what I have put in place so far (bearing in mind this is day +4 from writing my first hello world)

I need to create N copies of my system all initialised to the same values but each replica at a different temperature. The temperature of the replicas must not change. I then update each replica by a loop that advances the variables (this is a shared job between several processors). One the loop updating the variables is finished I need to exchange the variables to the next temperature and repeat the variable updating. The loop updating the variables is temperature dependent you see.

So my plan is, create N unions, all parametrised by different static global variables T_k. Initialise all the variables to the same value

then update them via a loop, when the loop is done, somehow exchange the updated variables between unions and then repeat the loop that updates the variables

@GlenH7 precisely :D

run metropolis-Hastings then exchange between replicas at adjacent temperatures and then iterate with Metropolis hastings

haha :)

haha thats a nasty list right there! I tried to drop organic chemistry and labs as soon as I could! Thank you for the help Glen!!

okay, so don't actually have N separate structures?

But I need to exchange between replicas at different temperatures

Quite a lot, it is more about a large set of variables than different molecules. I thought replica exchange monte carlo required there to be N copies in real time

@MichaelT the delete the directory solution? because that didn't work

Ah, No this is slightly different.

@Snowman NO! STOP.

Upgrading is bad. So many problems.

Do a fresh install.

Trust me.

Yes I have a large data set that can be represented as you say by a base structure and I will just in -put the data for each molecule type. But I think the method requires me to have N replicas of the system and then update each replica by standard Metropolis hastings then exchange between replicas then repeat MH

linux

well I am writing the code on my mac at the moment haha

but running it through linux

looks promising I'll give it a read!

signaling for what?

There is lots of complicating factors such as splitting this up between processors that I haven't even looked into yet.

ah, the good thing is that in the existing code that i am working alongside there are lots of parts that deal with how to split the job up over processors

(it deals with another MCMC technique unrelated to my Parallel tempering algorithm)

yeah .. I was planning on leaving that part for a while haha

I agree with both of you, I am confident that I can manage to split it up between processors without that much trouble. It's just this buisness of how to make N exact copies of a configuration space that is getting me

sounds promising

now that does sound promising!

so I can initialise them all the same way?

update independently via a loop and then exchange information between configuration spaces?

:D

thanks for your help Glen !! I really appreciate it :)

if you ever need help with any chemistry let me know haha ;)

So you are an engineer? What kind?

nice :) I did cell and molecular biology for a couple of years, Scottish universities require you to take another subject for two years! Really liked it though!

fair enough. I am in the middle of a bit of a career change. Been a theoretical chemist all the way so far but changing over to theoretical phys. Finding it tough but as you say such is life. ... gotta do what you enjoy doing ! :)

despite that my degree was material chemistry

Its all about the funding tbh. If you can get funding to look at something your set! I was working in a lab a little while ago and they lost the access to the main cluster :/ not great for a computational chem group

I've got a design problem that may be too broad for main

@Snowman I had nothing but trouble after upgrading from Win 7 -> Win 10. All the problems went away after I had Microsoft give me a key and did a total wipe.

here is the scenario:

I have network data feed -> third party library -> my own classes (adapter pattern)

I cannot make real instances of third party library objects without a data feed

in my tests, i can mock third party library objects, but if i am wrong about how they are structured, the test is useless

We actually discussed this but you can't allow all your data and work to be stored by a third party :/ You rely on their security and regulation of access etc

this means a problem can make it all the way to production before it gets noticed.

@GlenH7 that's not a unit test

My integration tests have... not been very thorough.

Hmm, they since got access to 150Tb cluster so I don't think they are complaining

maybe that's the problem, i need to have a more locked down process for integration tests.

Well, is there a main askable question here? if yes i'll just put it on main

well indeed, I left the group though so I don't know what they are up to anymore!

kinda think it's too broad though.

Ugh I wrote it up in main but it's sooooo broad

Alright I'll post it but if you think it's too broad don't waste your close votes, just tell me to delete it here and I will.

@Snowman okay i'll close it as duplicate too if you can find one

Hello everyone

(I'm the "noob" here)

plenty of noobs here!

I love it when I can eliminate a class that is a bunch of static methods.

I also love it when a plan comes together.

That's comforting, lol. I'm here pretty much at the recommendation from one of my not-so-successful questions asked on the forum

***question board

Yes

You'll get a lot of flak if you call it a forum. forums have very different expectations than a dedicated Q/A site

Yes, yes, I'm aware

My fingers simply decided to type the wrong word...lol

Anyway, my question is this one:

Here's the motivation behind my question:

I studied mathematics in college, and by the time my interests fully developed, my econ minor was done, and I had very little time (or money) to pursue a comp sci double major/minor

however, I'm now really interested in both computer science and in programming (yeah, I know they're not the same, do not worry)

after taking some object oriented programming courses, and learning a couple languages on my own,

what technologies are you referring to?

I've arrived to the question many get: "What's next?"

@daOnlyBG What's next? Yes. Yes is next,

and that led to, of course, "what makes a mature programmer mature? how do you like at a code and say: 'oh, this is novice's code,' versus 'oh wow, this is pretty good stuff'?"

@daOnlyBG you are kind of asking, "what bricks are examples of an experience bricklayer?" rather than "what makes an experienced bricklayer good?" - you can't just look at the bricks of coding and ignore how the whole structure works

you can have sweet bricks, but if you build the wrong thing or put them together wrong? that doesn't matter

I appreciate all these answers

Someone posted the "programmer's competency matrix" (sijinjoseph.com/programmer-competency-matrix) and I thought that was pretty close to what I was looking for

if anyone else wanted to build onto this answer/discuss it in the chat, I'd be interested to hearing what others have to say. that's all. :)

I appreciate the advice on "go code, and then code some more," and I can't disagree with that sentiment- I just worry if there's more to do than simply "coding more," as I wouldn't want to just keep using the same coding structures over and over again, and not end up learning anything

look through the standard library documentation so you know what wheels you have available

@MichaelT, thank you- that's a really cool index of topics I'd been curious about, but didn't know where to find comprehensive resources

Anyway, I'd like to thank you all for considering my question

I like to build small simulators in C#; I think you'll be seeing me here often, asking for help... so I thank everyone in advance

Oh, I did

I especially noted the differences in "team skills" among the multiple stages

Is it safe to assume that most competent programmers have a decent understanding of software architecture, or is that a more specialized field than I'm giving it credit for?

> I cannot generate real ones without a real data feed.

I even bolded it prior to your answer

Their library is pretty heavy weight as far as making a bunch of threads, checking a license key, etc., I do not want to start up their library in my unit test

which basically means that task 1 is the same as task 2

Now I'm not creating a "unit test", I should create an entire "test" program

i think it has to be the way to go, it's basically what the other answer is suggesting, it's a prototype

@Snowman The only way to do that is to spin up the entire library

Hi, Good Evening Everyone !

May I ask non technical question ? But important one

Currently I'm working on a Web based product company. I got a invite from another company which has Desktop based product. Should I take this chance? I wanted to know the aspects of technology wise - Thanks for all your opinions

I couldn't see a single JavaFx vacancy posted in SO Careers

I already have almost 2.5 years exp in WEB developing

@Snowman Nods. But is Desktops apps are dead now?

@MichaelT yes. correct. But after 2 years, if i work with desktop apps, and when I look for another good place, if no one using Desktop apps, that working experience will not be given a value. isn't it? Compared to if stay in WEB developing for those 2 years

@JudeNiroshan don't underestimate how slow large corporations move to update their technology, especially to web-based applications

so, this movement will not be actually a bad choice. isn't it? cambio.lk

this is the company which I got an invitation

@MichaelT coderanch.com/t/656270/Jobs/careers/…

will you agree with the first para in the last response ?

It is possible to use Spring in JavaFx ?

there is still plenty of cash money in desktop apps

Well, seems like it's actually not the way I thought of it.

@whatsisname yup. They pay more

there are still tons of apps written in straight win32 or MFC that need to be maintained that people will pay big money for

and even then, everything is ultimately moving memory around and doing math, you can don't have to choose desktop or web or mobile, you can do it all

@whatsisname yes. nice words

@Snowman hahahah. seriously?

seriously