Maths is hard

Quill

20:56

I'll tell you what, post a gist link in here and I'll take a look at it for you

TanMath

which is why I posted this to get code that is shorter to run so people on stack overflow could help me... but it is off-topic... therefore I am just going in circles

Vogel612

well you should fix one of those two things...

TanMath

@Phrancis have you even seen my question? it isn't people do often and post on the internet...

Vogel612

yourself or somewhere else :/

TanMath

@Quill you sure you have the time?

Quill

20:58

maybe, I can't be 100% until I see some code

TanMath

it is in the question:

1

Q: Open quantum system modelling

I have been working for a long time now on modelling an open quantum system using the Lindblad Equation. The Hamiltonian is the following: However, two other matrices are added to the Hamiltonian. One of them has all the diagonal terms equal to -33.3333i and everything else zero. Another is a ...

I will put it in a gist

Phrancis

Have you stepped through the math part of it, and compared the wrong result you are getting to the correct one?

Quill

can't you just reduce the list of matrices to reduce run time?

TanMath

@Quill what is the point of that? then how will I get the result I want??

Quill

well the calculation time depends on complexity, right? wouldn't reducing the complexity reduce the run time?

TanMath

21:01

gist.github.com/TanMath/93f5f7b8b4f7dccd00aa

@Quill but that would be a totally different system, wouldn't it?

Quill

I don't know the Lindblad equation, I'm just suggesting

Phrancis

options = Options(nsteps=1000000, atol=1e-10)

TanMath

@Quill yeah... you would get a different quantum system

@Phrancis what about that?

Phrancis

Any particular reason you are doing this a million times?

TanMath

@Phrancis um, because I want to plot the result correctly i guess... hides and runs away to check the code

Phrancis

21:04

Like, if you did it say, a thousand times, or a hundred, couldn't you still check the results then?

TanMath

@Phrancis I am checking if that affects anything right now...

Phrancis

Also, 1e-10 is a REALLY BIG number

TanMath

It has been a minute now, so it does take longer than a minute if I decrease it to 100 steps

@Phrancis you mean small?

Quill

how do you do this whole "Lots of tabs" thing @skiwi? it's really hard

Phrancis

1e-10=1*10^(-10)

(for my own reference, because I don't math)

TanMath

21:07

@Phrancis yeah.... so it is really small

code still hasn't finished...

Phrancis

No, it's a REALLY BIG negative number, it's not small

So, if you can cut that down by a few factors also, that should help I think

TanMath

@Phrancis headdesks

2

Quill

Phrancis: SQL guy by day... Bug Killer by night

Watch out bugs, he's coming for you!

Mat's Mug

@Quill and zombie hunter at times

TanMath

can somebody explain to @Phrancis that 1e-10 is a really small number?

Phrancis

21:10

How many zeroes in 1e-10?

Quill

3 mins ago, by Phrancis

No, it's a REALLY BIG negative number, it's not small

TanMath

@Quill it isn't even negative!!

Quill

Phrancis

OH

TanMath

^^^ @Phrancis

Phrancis

21:11

I'm sorry.

4 mins ago, by Phrancis

(for my own reference, because I don't math)

So, this 0.0000000001 figure, what is it for?

TanMath

@Phrancis maximum tolerance...

Quill

^ go here please

Phrancis

@Mat'sMug OK let me move this stuff over

Quill

first

The current discussion topic is the time-complexity issues on this script

2

Caridorc

hi, well I know pretty much nothing on numpy...

Phrancis

21:16

← 51 messages moved from The 2nd Monitor

Feeds

Quill has added Phrancis to the list of this room's owners‌.

TanMath

ok... so the program still takes forever to run...

oops.. I realized something... the trap matrix was all messed up... it should have been trap[2][2]=-0.03333j

@Phrancis qutip.org/docs/3.1.0/apidoc/functions.html#module-qutip.mesolve

Phrancis

@TanMath Would reducing the precision of this figure while you debug it help performance without screwing up the algorithm?

TanMath

qutip.org/docs/3.1.0/apidoc/classes.html#qutip.solver.Options

Quill

@TanMath try moving your variables into readable constants

like instead of trap[2][2]=-0.03333j

have:

trap_matrix = -0.03333j
trap[2][2] = trap_matrix

this will help you to be able to read and maintain your code easier

TanMath

21:24

@Quill what is the point of this? to make a new variable?

Quill

34 secs ago, by Quill

this will help you to be able to read and maintain your code easier

it makes it easier to find bugs in your code and to improve things

TanMath

maybe...

Quill

also makes life easier if anyone else ever wants to maintain your code

TanMath

@Quill ok...

@Phrancis well, the results would be less accurate...

Phrancis

Well, it might be at least worth a try just to see if that's the speed issue, maybe try making the tolerance just 1 and run it and see how long it takes

TanMath

21:26

wait... sorry!

Phrancis

?

TanMath

@Phrancis you are right.. atol should be higher, because then it considers it converged faster and takes less time that way... I will try that...

Quill

@TanMath time complexity is usually a byproduct of accuracy, unfortunately

TanMath

And by the way, please try to run it. Somebody who helped me with this code (@JRichardSnape from StackOverflow) actually got it to run in 10 minutes somehow... Maybe it is something with my computer...

@Quill i guess...

Only problem is you need to install the qutip module...

Quill

pip won't even let me install qutip

TanMath

21:30

@Quill what was the error? what are you using? Mac OS X?

Quill

osx:

Command "python setup.py egg_info" failed with error code 1 in /private/var/folders/5z/r7gx8xxs0zg8bc98tdxnkffh0000gn/T/pip-build-PGoEOQ/qutip

TanMath

it prefers Linux

Quill

right

TanMath

@Quill haven't seen this error before... do you have all the required modules?

Quill

probably not but usually it tries to install all the required modules

TanMath

21:32

qutip.org/docs/3.1.0/installation.html#general-requirements

@Quill I tried doing it on a Windows computer and it never worked.. it was because of the requirements... then I used a Linux VM and an online VM...

Quill

fair enough

I think I have ubuntu installed on here anyway

TanMath

good...

Phrancis

Back to the code?

starttime = 0 # this is effectively just a label - `mesolve` alwasys starts from `rho0` - it's just saying what we're going to call the time at t0
endtime = 1000 # Arbitrary - this solves with the options above (max 1 million iterations to converge - tolerance 1e-10)
num_intermediate_state = 10000

What are these for?^

TanMath

gtg... I will be back...

Phrancis

k

TanMath

21:35

thanks for helping me BTW! I appreciate it...

Phrancis

No problem. Just remember this is not "what we normally do" :)

TanMath

21:54

@Phrancis I totally realize that...

Phrancis

OK good

TanMath

22:10

@Phrancis how long to run the simulation

@Quill I hope you got qutip installed on your Ubuntu without any problems...

Quill

oh sorry I was busy with something else, but it looks like I uninstalled Ubuntu...

Phrancis

@TanMath OK, can you reduce those also while you debug?

TanMath

@Quill no need to say sorry to me...in fact, I should say sorry for taking up your time...

@Phrancis meaning?

Quill

@TanMath decrease these variables

40 mins ago, by Phrancis

starttime = 0 # this is effectively just a label - `mesolve` alwasys starts from `rho0` - it's just saying what we're going to call the time at t0
endtime = 1000 # Arbitrary - this solves with the options above (max 1 million iterations to converge - tolerance 1e-10)
num_intermediate_state = 10000

Phrancis

What Quill said. There's no point running a long simulation if you know the result is likely to be wrong anyways, just run a short simulation until the numbers look accurate enough

Also, floating point calculations are a real PITA lol

TanMath

22:18

@Phrancis what is PITA?

Phrancis

Pain In The Ass

TanMath

Oh...

Phrancis

But if you change endtime say from 1000 to 100 or even 10, can you still get enough of a result set to see if it's wrong or if it looks OK?

TanMath

@Phrancis probably... I will set it to 100 maybe...

Phrancis

And what about num_intermediate_state what is that for?

TanMath

22:24

I will make num_intermediate_state=100 now...

Phrancis

(Need to use backtick ` to get the monospace formatting)

Transcript for