The h Bar

Terry Bollinger

17:00

@0celo7 teeth, not nanotubes :)

DanielSank

@DavidZ I tried to explain this in that meta post about biophysics. The summary is that it's easy to look at a bad post and say "this doesn't belong here".

Terry Bollinger

Tis noon, bye all. No color experts out there today!

DanielSank

Saying "I don't understand what you're asking" or "too broad" makes me feel like I should explain how to improve the post.

But if I say "off topic" then I feel like I'm done.

I wonder if anyone else feels the same way. I suspect so, because there are a lot of posts closed as "off-topic" where I think "too broad" or "unclear what you're asking" make way more sense.

David Z

@TerryBollinger bye!

ACuriousMind

I think @DanielSank has a point - there might be a tendency to vote stuff as "off-topic" when the question is really so muddled that one can, strictly speaking, not tell whether it is off-topic or not.

2

DanielSank

17:01

^ YES

That's what I'm trying to say.

David Z

Ah, a good point.

John Rennie

@ACuriousMind Yes I sometimes do this. Don't we all?

David Z

Not much we can do about it, though

DanielSank

I think this could be improved. Saying "off-topic" tells OP to go elsewhere. We provide no information on how to ask better, which is almost always what OP needs.

@DavidZ Oh, I think we can!

David Z

@DanielSank well, that's exactly what we were trying to do with the homework policy being linked in the close reason, and presumably the same would happen with a new insufficient effort or calculation request close reason

@DanielSank like what?

DanielSank

17:04

@DavidZ 1) change the wording of the homework close reason so that it's more clear where it should apply. This is an ongoing discussion, obviously.

ACuriousMind

@JohnRennie Yeah. But this makes discussing the off-topic policies a bit harder. Just because we see a lot of crap posts of type X, this doesn't mean X should be off-topic (it just means crap posts should be closed for being crap, not for being about X)

DanielSank

2) In the off-topic reason, specifically say to use it as a last resort.

3) Put the off-topic reason lower down in the list relative to the others.

David Z

@DanielSank ok, well, let's operate under the assumption that we're scrapping the homework close reason and developing a new one from scratch. I'd want to know how to change the wording to make that more clear, so that we can incorporate those changes in the new reason.

By the way, the chat session time is over. We'll keep discussing, but see everyone in two weeks when we will hopefully have more to say about the replacement for the homework policy!

DanielSank

I'm trying to point out easier stuff first.

David Z

Fair enough

DanielSank

17:06

I think we're over-using off-topic not just on homework, but generally on bad posts.

See what I mean?

0celo7

I agree.

DanielSank

Unfortunately I need to go to work. I'll try to chime in some more later.

ciao

David Z

@DanielSank that we cannot do - I mean, we can probably change the order relative to other custom off topic reasons (engineering and the other one), but not relative to unclear or too broad

Thanks @DanielSank

see you later

DanielSank

@DavidZ Oh, that's too bad. Oh well.

David Z

@DanielSank yeah, I think that's true, and that's ~~part~~ most of the reason we're replacing the homework-like close reason in the first place

John Rennie

17:10

How about we change homework-and-exercises back to just homework, then add an insufficient-effort close reason?

David Z

@JohnRennie That would imply that we're able to distinguish homework from exercises

which we really can't

John Rennie

Exercises are an important part of learning.

I don't see the problem with explaining how a worked exercise is done iif I believe the OP is genuinely interested and not just lazy

0celo7

@yuggib Do you have a recommendation for a PDE book for a beginner that doesn't require functional analysis and measure theory and all that? Preferably Springer so I can easily supplement the assigned textbook.

David Z

Yes... well, okay, what I meant was that splitting the tag implies that we know how to distinguish homework from non-homework, and furthermore that we care about the difference. Which I think we shouldn't.

0celo7

@JohnRennie How does the OP convey this interest

David Z

17:13

@JohnRennie I guess the question with that is, how accurate is your sense of whether the OP is interested or just lazy?

John Rennie

@DavidZ and @0celo7 it's a judgement call. But is it a catastrophe if I'm not right 100% of the time?

0celo7

@JohnRennie Probably

David Z

No, but what about if you're right 10% of the time? or 50%? And similarly for everyone else who is trying to make that judgment?

0celo7

0

Q: Derivation of Schwarzschild metric using the full machinery of differential geometry

How would one derive the Schwarzschild metric using the full machinery of differential geometry, using the component approach as little as possible?

ACuriousMind

@JohnRennie Not if the criterion was reasonably objective. But "conveying interest" is a) terribly subjective and b) more about the asker than about the question. We should not be closing questions for inferences we make about their authors.

0celo7

17:16

@JohnRennie Off-topic?

David Z

@0celo7 that's actually a good case to consider for our new policy

John Rennie

@0celo7 well it doesn't show a lot of effort does it?

No indication what the OP has already found out about the area

The OP gives no indication what they mean by full machinery of differential geometry

Actually I'd be tempted to VTC as unclear

It isn't obvious to me what sort of answer the OP wants

0celo7

@ACuriousMind Should there be another star here? $\nabla\star j$ is not a scalar. (Also since when do you use \ast for Hodge?)

John Rennie

How about:

0

Q: Reference frame and conservation of energy

Say spaceship $\alpha$ burns a portion of its fuel to leave planet A and is cruising through space at 10 m/s relative to the surface from which it launched. Spaceship $\alpha$ is being observed by spaceship $\beta$, which launched from planet B. The relative velocities of the two planets are such...

newtonian-mechanics energy energy-conservation reference-frames relative-motion

0celo7

@JohnRennie I just did...cf. my comment

ACuriousMind

17:22

@0celo7 Yes. (I used it because I forgot what one had to do with the star to make the spacing not horrible)

John Rennie

This looks at first glance like a homework problem, but actuall it's a conceptual failure.

You can answer the problem by explaining what the OP has misunderstood without actually doing the calculation for them

0celo7

@ACuriousMind If $\star\nabla\star j$ is $\nabla_aj^a$, what is $\star\mathrm{d}\star j$

I can never remember this

ACuriousMind

@0celo7 $\partial_a j^a$

0celo7

@ACuriousMind BS

that's not a scalar

ACuriousMind

wat

0celo7

17:23

$\star\mathrm{d}\star j$ is a scalar, right?

ACuriousMind

That's the divergence, and the divergence operator is ${\star}\mathrm{d}{\star}$.

0celo7

On a general manifold $\partial_aj^a$ is not a scalar.

ACuriousMind

I have a feeling we've had this debate before

0celo7

We have!

It's not a debate, it's me asking how a scalar can equal a non-scalar ;)

ACuriousMind

Ack, damn

${\star}\nabla{\star}$ is not the covariant divergence, ${\star}\mathrm{d}{\star}$ already is.

0celo7

17:27

I know :)

Let's reverse the question

What is ${\star}\nabla{\star}j$

I don't wanna calculate it

David Z

@JohnRennie it looks like exactly the kind of homework question we want - or I might say, it looks like exactly the kind of question we want, and I don't really care whether it's motivated by homework or not

0celo7

@ACuriousMind Sigh...you really don't get abstract index notation

$a$ is an abstract index, so summing over it makes no sense

it should be $\nabla_a j^a:=\sum_\mu\nabla_{e_\mu}j^\mu$

ACuriousMind

@0celo7 $\mu$ is a dummy index, I might as well call it $a$ :P

0celo7

@ACuriousMind but then you confuse HE/Wald fanboys :(

ACuriousMind

@0celo7 I...don't think I disagree :D

0celo7

17:34

I need to write a GR book to set the whole index thing to rest

Mike Miller

yeah, one without indices

0celo7

Early Latin for abstract indices, middle Latin for spatial coordinate indices, early Greek for coordinate indices in SR/flat indices, middle Greek for spacetime coordinate indices, capital Roman for abstract spinor indices, capital Greek for spinor coordinate indices

Something like that

ACuriousMind

"early" and "middle"?

0celo7

abc vs. ijk

ACuriousMind

You mean, like a,bc, and k,l,m?

Ok

0celo7

17:37

$\alpha\beta\gamma$ vs. $\mu\rho\sigma$

ACuriousMind

@MikeMiller It's...actually pretty difficult to translate all the physicist's indices into coordinate-free notation.

And you never get the prefactors right

0celo7

^ Straumann tries.

Mike Miller

@ACuriousMind: I have a solution.

0celo7

But even he gives in at points.

@ACuriousMind Prefactors?

@MikeMiller Stop doing physics?

Mike Miller

: - )

ACuriousMind

17:39

@0celo7 Like, is that $\frac{1}{2}$ over there subsumed in your convention of whether the components of $\omega$ are $\frac{1}{n!}\omega_{1\dots n}$ or is it "actually" there? Especially when objects with more than two indices are floating around, it can get pretty confusing which of those factors remain in the coordinate-free notation.

0celo7

@ACuriousMind Oh, right.

Also how does coordinate free notation deal with $R_{abcd}\mathcal{R}^{acdeklm}D^b{}_eC_{klmn}A^n$

I'm sure that's a thing in someone's GR paper

Mike Miller

A related problem on the math side: There are two ways of defining $\alpha \wedge \beta$ when these are $\mathfrak g$-valued forms, that differ by 1/2, and you have to figure out which it is from context. They're otherwise written the same.

This is even though most authors agree that when they're $\Bbb R$-valued forms, there's no 1/2 factor.

0celo7

@MikeMiller why is that again?

I've never dealt with forms that aren't $\mathbb{R}$-valued

ACuriousMind

@0celo7 Liar, the Christoffel symbols are $\mathfrak{gl}(n)$-valued forms ;)

Mike Miller

The structure equation gives that if $\omega$ is a connection form and $\Omega$ its curvature, then $d\omega = -1/2 \omega \wedge \omega + \Omega$. Some people don't like that 1/2.

Other annoyances: that term is sometimes written $[\omega,\omega]$ and $[\omega \wedge \omega]$. I think the first one excludes the 1/2 and the second usually includes it. They're all dumb.

DanielSank

17:54

@ACuriousMind That. That is the critical point.

I don't care at all about the user's mental state when they post a question. I only care about the quality of the question itself.

Another well-put statement by @ACuriousMind.

@JohnRennie Good, "this is unclear" is more useful than "this is homework".

0celo7

@ACuriousMind and the spin connection is $\mathfrak{o}(n)$

whatever

smartass

Mike Miller

That's not really being a smartass, connection forms are basically the place one cares about $\mathfrak g$-valued forms.

DanielSank

@ACuriousMind can you give me a mental/geometric understanding of the Hodge dual?

ACuriousMind

@DanielSank It sends a $k$-hypersurface to the $(n-k)$-hypersurface orthogonal to it. Like, in three dimensions, it sends a plane to its normal vector (that's what the cross product does - given two vectors, it gives the one orthogonal to both)

user54412

@0celo7 yes

0celo7

18:08

@MikeMiller oh, I've seen that

hubot

What is relations between physics and psychology?

0celo7

18:31

@ACuriousMind Huh, I've never seen it put that way

How does one see this?

ACuriousMind

@0celo7 Just by its definition. For an orthonormal basis $e_i$, we have ${\star}(e_{i_1}\wedge\dots\wedge e_{i_k}) = e_{j_1}\wedge\dots\wedge e_{j_{n-k}}$ where the $j_{\dot{}}$ are precisely the indices not among the $i_{\dot{}}$.

0celo7

18:52

@ACuriousMind oh, yeah :P

@ACuriousMind So what is $\star\nabla\star j$ anyway

Bernard Meurer

@DanielSank Would you know why so many people seem to 'godify' functional languages?

Bass

@ACuriousMind Cool, didn't know/realize that. Is that the Ehresmann connection stuff?

ACuriousMind

@0celo7 dunno

@Bass No, an "Ehresmann connection" is phrasing a connection in terms of choices of horizontal subspaces

Bass

@BernardMeurer Do you like SQL?

Bernard Meurer

@Bass I do yeah, I think it's great at what it does. I don't love it like Python, but I enjoy it.

DanielSank

18:58

@ACuriousMind Excellent. That makes perfect sense. Thank you.

If I provide a set of $k$ vectors which define the hypersurface, is there a simple explicit formula which generates the $n-k$ vectors which define the orthogonal one?

I mean, this is what the cross product does, so I suppose there's some determinant-like thing which gets the general case right.

@BernardMeurer Well, I think it's because they're so simple.

There's no state.

ACuriousMind

@DanielSank Well...the formula is given by the formula for the Hodge dual...

Bass

@ACuriousMind Maybe what I mean makes sense only in principal bundles. In a principal bundle, the vertical subspace of the tangent space is $\mathfrak g$, right? So the Ehresmann connection is a projection $\omega:T_pE\to\mathfrak g$. At least that's how I understood Nakahara so far.

Maybe it's just some random association, I read "$\mathfrak g$-valued form", and thought of that.

DanielSank

@BernardMeurer Furthermore, functional programming more or less forces you to write composable code. You are forced to describe your problem as a pipeline of information. Each function represents one step of the pipeline.

Danu

@ACuriousMind Little question on ST, see if you know: If I've got a vertex operator $:e^{ik\cdot X}:$, how do I explicitly do the normal ordering?

DanielSank

Because of this, it becomes commonplace to write code in a way that facilitates re-using composable pieces.

Danu

19:05

$:e^{ikX}:=(1+ikX-(kX)^2/2+\dots)+$contractions

DanielSank

The stateless aspect is really important. It's just really difficult to do something stupid if you don't have to keep track of the state of some field in a class etc. etc.

Danu

but the contractions are strange, because the $X$'s are evaluated at the same point so for e.g. the second order term I'd get $+\frac{\alpha'k^2}{4}\ln(z-z)$

DanielSank

I'd encourage you to try out Scala, because it provides language features for writing functional code, but also lets you do things the object oriented way whenever you want.

ACuriousMind

@Bass Ah, well, yes, every connection form is that projection. One doesn't talk in those terms in usual (pseudo-)Riemannian geometry, though, and there's a special thing - the principal bundle one might consider here is the frame bundle, and it comes with a solder form.

Danu

...and I don't like evaluating $\ln$ at zero

Bass

19:07

@DanielSank One more thing that many people see as an advantage of functional languages is that they are more declarative than imperative. You don't tell the computer what to do, you declare some structure or logic, and the computer has to find a way of evaluating that.

SQL has many functional/declarative aspects, for example.

@BernardMeurer you could also have a look at the 101 LINQ samples. LINQ is a combination of O/R-mapper and functional language inside the .NET languages C# and VB.NET

@ACuriousMind Gonna read about the solder form, never heard of it. Thanks.

NikolajK

They may also sometimes permit an overlying type system that expresses dependent function types using fibrations :)
(because you were speaking about fiber bundles)

ACuriousMind

@Danu What do you mean "explicitly do the normal ordering"? You essentially always use Wick's theorem to compute the OPEs of $:\mathrm{e}^{\mathrm{i}kX}:$ with stuff

Danu

Say I just want to write out $:\exp ikX:$ explicitly in terms of the $X$'s in an infinite series

ACuriousMind

Yeah, that's gonna be singular

Danu

Yeah

ACuriousMind

19:19

Why would you want to write it out like that?

Danu

So how do I justify that for an OPE involving this thing, I don't have to take the "inner contractions"?

Bernard Meurer

@Bass VB.NET gave me a mini stroke last year

@DanielSank I'm excited to try Scala out, any books/sources you recommend?

Bass

@BernardMeurer You should have tried out C# instead, it's nice. VB.NET is horrible.

Bernard Meurer

@Bass People have been telling me to try out C# for ages (not so long considering my age) but I just never quite got around it. I never felt the need to do something I could not achieve with C or Python that caused me to go towards C#

TanMath

@DavidZ sure...can you please start from this message?:

20 hours ago, by TanMath

@DanielSank here is the PEP8 version:

Mark Mitchison

19:25

@TanMath I'm not ignoring you, I just haven't been online much

I have to say that I'm not willing to read through your code, though

so if it is a coding problem, I can't help

TanMath

@MarkMitchison oh...I keep seeing you on the chat...

so I was confused...

Mark Mitchison

I would suggest a simple physical check on your results

TanMath

@MarkMitchison I believe it isn't...

Bass

@BernardMeurer If you know C and Python, you have no real need for C#, that's right. If you'd like to see why functional languages are hyped, take a look at LINQ. There's even a port of LINQ to Python, Pync. Don't know whether it's good/stable/active.

Mark Mitchison

for example, how does the trace of the density matrix change in time

TanMath

19:26

@MarkMitchison how would that help?

Bass

Scala will do, too.

Mark Mitchison

i.e. the sum of the populations

ACuriousMind

@Danu Oh god that's horrible to do right. You have to take them, it just turns out they "cancel" in the appropriate way. Formally, you expand the exponential, thenwrite something like $X(z)X(z)X(z)$ as $X(z_1)X(z_2)X(z_3)$, chug through the calculation, and take the limits $z_2\to z_1$ and $z_3\to z_1$ in the end. If you've done everything correctly, the result will be fine

Mark Mitchison

@TanMath well, what do you think the trace (sum of site populations) should do over time?

TanMath

I would like to note that the current plot of the site 3 population looks like the site 2 population with no loss or trapping...

@MarkMitchison stay constant?

Bernard Meurer

19:27

@Bass I wanted to go into Scala because Java is my ultimate nemesis

Mark Mitchison

@TanMath No. You have anti-Hermitian terms in the Hamiltonian

The loss terms

John Duffield

@Vogel612 : all points noted Vogel.

TanMath

oh yeah...

Mark Mitchison

These should cause the total site population to vanish eventually

TanMath

yes...

Mark Mitchison

19:28

But you do not see loss in your results. Hence I would do some simple check to see whether the loss terms are really behaving properly

such as calculating the sum of the populations

TanMath

so that will help me determine if the problem is with the trapping...

Mark Mitchison

You could also just remove the anti-Hermitian terms and see what happens

ACuriousMind

@Danu: Rather, you have to "understand" that $:X(z)X(z):$ is really defined as $\lim_{z_2\to z}\left(R(X(z)X(z_2)) - C(X(z)X(z_2))\right)$ where $R$ is radial ordering and $C$ is contraction

TanMath

@MarkMitchison I added the trapping hamiltonian to the excitonic hamiltonian...

@MarkMitchison actually, before, when I forgot those terms, it produced the same results...

ACuriousMind

Both the radially ordered terms and the contraction diverge at $z_2\to z$, but their difference is finite.

TanMath

19:30

but I do not understand why it would not work...

Bass

@BernardMeurer Scala compiles to Java bytecode, what's the point?

Mark Mitchison

@TanMath Well that seems pretty strong evidence that the loss terms are not behaving properly

@TanMath Yes, because you do not understand how the code you are using works

This is a problem

It might be that QuTip simply cannot handle a non-Hermitian Hamiltonian

TanMath

@MarkMitchison so it is a problem with the code, and more specifically with the trapping term?

@MarkMitchison I am pretty sure it can... it was meant for open quantum systems!

Mark Mitchison

@TanMath I don't know! It's your job to find out. I'm just giving you some suggestions.

TanMath

ok...

Mark Mitchison

19:32

@TanMath You are using a function that simulates a Lindblad equation. Strictly speaking, the Hamiltonian should therefore be Hermitian. The non-Hermitian Hamiltonian means that the equation is not trace-preserving and therefore cannot be a Lindblad equation by definition.

Bernard Meurer

@Bass The point is learning something that runs on a java VM other than Java that causes me to cry on the floor

TanMath

@MarkMitchison wait, what??

so it isn't a lindblad equation?

Mark Mitchison

he Lindblad equation is the most general completely positive, trace-preserving equation of motion for a density matrix

Look, this is a matter of definition.

I don't know what the writers of QuTip thought

But to just assume that you can put non-Hermitian Hamiltonians is not justified

TanMath

yeah... I probably just forgot the definition... you are right... I did not think about that...

Mark Mitchison

Unless it specifically allows for that in the documentations

TanMath

19:35

@MarkMitchison now, why would my model have a non-hermitian hamiltonian and the authors claim to be using the lindblad equation?

Mark Mitchison

Like I said, it's a matter of definition

TanMath

is there a name of the equation I would be using right now?

Mark Mitchison

No

TanMath

because Lindblad is a generalization of Bloch-Redfield, correct?

Bass

@BernardMeurer You'd still be dependent on the JVM by Oracle, no?

TanMath

19:36

(sorry, I do not have my open quantum systems textbook!)

Mark Mitchison

@TanMath No, the Redfield equation is not necessarily a Lindblad equation,

Bernard Meurer

@Bass, on the JVM which I am fine with, but not on the Java language itself and its obnoxious NullPointerExceptions ~shivers~

Mark Mitchison

@TanMath First you need to check that the non-Hermitian terms are not behaving properly. E.g. by calculating the trace

TanMath

@MarkMitchison then what is the relationship?

@MarkMitchison ok...

Bass

@BernardMeurer NPEs are present in other languages too, are they special in Java?

Bernard Meurer

19:41

@Bass Well, I'd say they are as bad as Segmentation Faults in C but they blow more things up when they happen in Java. Also, there's something about Java syntax that makes my neurons stab each other

Bass

@BernardMeurer I suppose the NPE handling is similar in Scala, since it compiles to the same bytecode. But yes, Java syntax has its weak points :)

Danu

@ACuriousMind Yeah, I know that

Thanks :)

Bernard Meurer

@Bass Don't wreck my dreams just yet

@Bass I don't even like C to be honest, I only use it because ain't nobody got time for letting Python calculate factorials.

TanMath

@MarkMitchison you there?

Mark Mitchison

yeah

TanMath

19:47

oh...

do you know the relationship between redfield and lindblad?

Bernard Meurer

@DanielSank Have you ever messed with Python Lambdas?

TanMath

@BernardMeurer oh, lambdas?

Bernard Meurer

@TanMath Yep, lambdas, have you?

TanMath

@BernardMeurer yeah...they are neat but have this srypid annoying little syntax...

Bernard Meurer

@TanMath I was looking for someone with some time to give me some tips on how to use them, I'm having trouble getting what they are/do

TanMath

19:53

@BernardMeurer they are simply anonymous one line functions...

Bass

@BernardMeurer Lambdas are pretty much the core of LINQ. So if Python has them natively, it is a (partly) functional language. Try to write some code just using lambdas (filter, aggregating, mapping), that's functional programming.

filter(lambda x: x % 3 == 0, [1, 2, 3, 4, 5, 6, 7, 8, 9]) filters the list to return the multiples of 3

TanMath

@MarkMitchison also, how can I perhaps convert my model into a different equation that can be solved?

Bernard Meurer

@Bass I deal a lot with data mining currently, and I have one massive class in a code called Filter filled with different filters, I thought something like lambdas could help me filter data sets in a cleaner way

@Bass huh that resembles list comprehensions a bit doesn't it?

Bass

@BernardMeurer Yep, IMHO one of the best features of functional programming.

user54412

@BernardMeurer yes, exactly

Bass

19:57

@BernardMeurer Isn't list comprehension just a way of declaring a list? FP is much more.

Bernard Meurer

So analyzing that code, the lambda x: x % 3 == 0, [1, 2, 3, 4, 5, 6, 7, 8, 9] is returning a set of multiples of 3, so what is the filter for?

user54412

list comprehension is basically a lambda with the input to the function all bundled into one very pythonic line

Bernard Meurer

@Bass you can declare a list with list comprehensions, but you can do some pretty bad-ass filtering/normalizing with it as well

@ChrisWhite List comprehesions are my favorite thing in python, it's such an elegant solution for so many problems

You can create a simple cypher in a single line with it

Bass

@BernardMeurer filter is like the WHERE in SQL. It's a function that takes two parameters: a list/array and a predicate. A predicate is a (lambda) function that returns true or false for a given item.

Danu

Is this Programming Chat now? :P

Bernard Meurer

20:01

@Danu I'm sorry if I caused a subject escapade, I just found that physicists are better explainers than computer scientists on average

user54412

@Danu Sorry, were you looking for obscure French cinema chat?

Danu

Oh, that's fine :) Don't worry!

@ChrisWhite hahaha, dat hostile response doe

user54412

:p

Danu

"obscure" is not what it is, though maybe for you 'Muricans (cough uncultured swines cough)... :)

Bernard Meurer

@Bass I did some messing on the interpreter here, and I think I get the Idea

Wow these things are damn cool

So in a functional language this kind of elegance would extend over to many other aspects of the language?

0celo7

20:09

@Danu wow bully

Bass

@BernardMeurer Well I think you can write everything strictly functional, if you really want. But that might well be a huge pain, since there are things that are lesser suited for it. Nobody writes low-level 3D rendering code in functional languages, I guess.

But there are many tasks in programming that occur very frequently. Mapping. Filtering. Sorting. Grouping. Joins. These types of things are quite elegant, fast to type and easy to read in functional langauges.

Certainly better than in purely imperative ones like C, where you quite have to type out everything.

0celo7

@ACuriousMind Does the exterior covariant derivative only act on tensor-valued p-forms

Bernard Meurer

@Bass I see what you mean. I just feel like I need to add something else under my belt. I'm going for college next year and Knowing C and Python I feel it would be nice to know a functional language. Also doing it in the hiatus before I leave brazil would give me the proper time to learn.

Bass

@BernardMeurer I don't know Scala, but I have heard good things about it. Go for it :)

0celo7

@ACuriousMind cause I think $\star\nabla\star j=\star\mathrm{d}\star j$

Bernard Meurer

20:15

@Bass Will do, just need a good book now

Bass

@BernardMeurer One of the coolest features about LINQ is that this functional syntax you just saw a glimpse of, can be translated into SQL statements on a relational database, or to other queries against some data source.

Bernard Meurer

I also don't know if I should go for Haskell or Lisp, so many options

Bass

The important thing is that when you program functionally, you first build an expression tree, an in-memory structure that can be used for many things. It can be compiled to machine code, like in the example you just did. It can be translated to SQL. It can be inspected, reflection-like.

@BernardMeurer Maybe first try to write some functional Python code to get a feel for it. Then, try out Haskell and Scala. Lisp is cool because it's so unique in its syntax, but I wouldn't want to work in it.

Bernard Meurer

@Bass I stopped learning Lisp because it broke my () keys on the keyboard

Bass

@BernardMeurer exactly :)

Bernard Meurer

20:21

I'll give it some good thought. Scala is on top of my list because it runs on JVM, Haskell because it's cool for AI I've heard

0celo7

Hawking warned against AI

Bass

@BernardMeurer Lisp has the reputation of being good for AI too, because everything in Lisp is a list, and most of its commands are list-manipulation, so there's some notion of learning self-manipulation of the program code. But maybe that's more cliché than reality :)

TanMath

@0celo7 that is the first dumb thing hawking did...

Bass

Gotta check out list comprehension. Looks cool.

Bernard Meurer

@Bass List Comprehensions are beautiful.

Now that you mentioned I do recall talking to a guy working with AI at IBM that used Lisp

NikolajK

20:31

In 30 years, doing theoretical physics and programming will be the same thing :)

John Duffield

@NikolajK : I doubt it.

NikolajK

@JohnDuffield: Me too.

John Duffield

Ah, I have just noticed the smiley.

NikolajK

That was a hopeful smiley.

0537

@NikolajK 100 years then?

0celo7

20:43

@ACuriousMind I'm going to write an answer on that abstract index question...hopefully I can teach you something!

NikolajK

There is a lot of work going on in pushing the Curry-Howard correspondence for a more expressive higher order logic, but the people developing stuff are mostly interested in content too rigorous for physics

https://en.wikipedia.org/wiki/Curry%E2%80%93Howard_correspondence

0celo7

it's very convoluted though

there's a reason why only Wald uses abstract indices

0celo7

20:57

god why does anyone use this notation

0celo7

21:09

@ACuriousMind Well...that answer is a mess

You should let me know if that makes things more or less confusing, I actually don't know

ACuriousMind

@0celo7 Less confusing, definitely.

0celo7

@ACuriousMind Cool

Now you understand abstract indices!

And why no one uses them!

@ACuriousMind Straumann says something like: I'll just use Greek indices throughout and they're abstract when they ought to be.

Mark Mitchison

21:27

@TanMath I have an idea which might work

If you enlarge the Hilbert space to include an additional state

The state with no excitations (vacuum state)

Then the loss and recombination terms can be represented as Lindblad dissipators

with Lindblad operators |3><vac| and its Hermitian conjugate

where |3> is the state with an exciton on the site where loss and recombination acts

and |vac> is the vacuum state

This removes the need for anti-Hermitian terms

(this is assuming that the loss and gain terms only act on site 3)

DanielSank

@BernardMeurer Sure all the time.

Btw, Scala's syntax for the same thing is a lot nicer ;)

Bernard Meurer

@DanielSank Cool! Would you be kind enough to recommend me a book if you know any?

DanielSank

@BernardMeurer Yeah for sure.

So, learning the language isn't that bad.

Try Scala By Example.

The tricky thing, as always, is learning how to actually get the code to build.

Bernard Meurer

@DanielSank this one?

I love bug fixing hahaha

I'm weird like that, my favorite part is making the broken things work

alarge

@BernardMeurer As mentioned before, functional languages are popular for their simplicity (in particular in that you lack a state), which makes e.g. testing very easy. They also make verification and proving that your program actually does what it is supposed to possible, and have direct mapping to type theory, category theory and all that mathematical jazz, "proper" theory of computation.

Bernard Meurer

21:42

@alarge Interesting, I like the mathematical aspect of that. Would you be able to say something about the general runtime speed of Scala or any other language? As compared to Python for example?

TanMath

@MarkMitchison the trapping only acts on site 3, but recombination acts on all 7

hey @DanielSank

0celo7

@ACuriousMind You probably won't know the answer to this, but: is it legitimate to write $\Gamma^a{}_{bc}$, i.e. the connection coefficients in abstract indices

I really can't decide

TanMath

@MarkMitchison what would that vector be?

ACuriousMind

@0celo7 You're right, I don't know the answer to that, but I'd guess not - the indices "label the type of tensor" (don't they?) and that's not a tensor

TanMath

all zero?

ACuriousMind

21:45

@TanMath "all zero" is not a state vector. You should know that a vacuum vector is not the zero vector.

alarge

@BernardMeurer Scala boasted accelerating some CFD code a couple of years back (I think Odersky, the designer of the language, mentioned it on the Coursera course he teaches). No idea what their benchmark was coded in. As for Python, it all will depend on the flavour (e.g. PyPy) and for performance you'd probably be calling libraries anyway.

TanMath

@DanielSank I hope you saw the chatroom...

0celo7

@ACuriousMind Yeah...but what's stopping me from writing $\partial_a g_{bc}$, etc.

user54412

I feel abstract indices were invented by some physicist being told "no, you're confusing tensors with their components" and responding "actually, these here are abstract indices, so I'm not wrong."

0celo7

@ChrisWhite It's just so confusing

Me writing that answer confused me

TanMath

21:46

@ACuriousMind i should know that...my brain isn't doing well today...

DanielSank

@BernardMeurer No no I mean understanding how to even specify dependencies, etc.

0celo7

maybe we have to stipulate that $\partial_a$ only has meaning on $C^\infty(M)$

DanielSank

@TanMath Stop pinging me for heaven's sake. I can read. I don't need your repeated pings.

0celo7

@DanielSank

5

TanMath

ok...

^

Bernard Meurer

21:47

And my favorite

0celo7

Don't worry @TanMath I'll ping him for you

TanMath

thanks!

0celo7

@BernardMeurer uh, thanks?

@ACuriousMind Fun fact: Wald literally says that $\Gamma^a{}_{bc}$ is a tensor.

ACuriousMind

@0celo7 Uh...are you allowed to write that? For $\partial_a f$ I see it's fine because the covariant derivative of $f$ is just the partial, but is $\partial_a g_{bc}$ really allowed?

Bernard Meurer

@alarge@DanielSank, What abou library availability? It's one of my favorite things about python; how does Scala relate to that?

0celo7

21:48

@ACuriousMind According to Wald it is!

TanMath

I am confused what he wants me to do...

ACuriousMind

@BernardMeurer Didn't take you long to catch that one ;)

0celo7

He's the one who invented the notation

ACuriousMind

@0celo7 Fun fact: Wald is wrong :D

TanMath

he says, make functions out of my code...

0celo7

21:49

@ACuriousMind You're telling me?

TanMath

I told him my idea, but he never responded...

alarge

@BernardMeurer I've never done anything serious with Scala or Java, but to my understanding you can call Java libraries which are many.

ACuriousMind

@0celo7 I know you knew that

0celo7

@ACuriousMind Please remind me why it's not using only the definition of tensor as a multilinear map

No coordinate transformation nonsense

Bernard Meurer

@ACuriousMind Being here a couple days things I discovered:
@0celo7 is a troll
@DanielSank answers questions nicely, but you better follow PEP8
@TanMath doesn't like PEP8
Some moderator doesn't like it when I say *abuse*
@alarge gives nice tips

0celo7

21:51

Or let's just look at $\partial_a g_{bc}$

Bernard Meurer

@ACuriousMind Knows almost everything and gives canonical answers mostly

0celo7

I'm reading Wald and now I've forgotten all math

ACuriousMind

@0celo7 Let me carefully think about that one before I respond

@BernardMeurer You've got me confused with @Qmechanic there ;)

0celo7

@ACuriousMind Just the derivative of the metric please. Where $\partial_a g_{bc}:=\frac{\partial g_{\mu\nu}}{\partial x^\rho}blah$

you can fill in the rest with your imagination~~

@BernardMeurer I'm not even a troll

Bernard Meurer

@ACuriousMind, never seen @Qmechanic here so I can't speak about him yet hahaha

Qmechanic

21:54

@ACuriousMind :)

Bernard Meurer

@Qmechanic what is the magic number for the Fast inverse square root algorithm when using IEEE 754 floating point format?

TanMath

well then...my questions aren't answered...I have no ides what to do...I just know thst I have to calculated and plot the trace of the desnity matrix...but MM and DS have nothing else to say...

so with no reason to stay, bye!

Bernard Meurer

Auf wiedersehen @TanMath!

0celo7

you a Kraut now

Transcript for