The h Bar

enumaris

19:00

arg

danielunderwood

Yeah I'm pretty cynical after seeing the inside too

Slereah

Hopefully they only do the dealing with clients softwares :p

danielunderwood

My view is that consulting companies don't really care about their software and the real product is sales...I don't know how accurate that that is though

vzn

← can extremely outcynical anyone :P

danielunderwood

It has made me avoid applying for that type of company in my current job hunt

enumaris

19:02

I dunno man...I just interviewed for a A.I. consultancy...I can't trash talk consultancies until they reject my application lol

danielunderwood

I'm kind of curious how AI consultancies work...do they just work on easy stuff?

enumaris

who knows

danielunderwood

If they didn't, I feel like they'd have a number of things where they just can't get a working model

Slereah

Most AI stuff fir companies is easy

Just the same types of problems

Look for patterns in sales and whatnot

enumaris

I bombed the anomaly detection question today...brain no worky at 7am-y

I learned anomaly detection before...but like 12 months ago...so...it didn't all come back to me until after the interview was over looool

danielunderwood

19:06

Probably better than my response of "anomaly detection?...uhhh"

The closest thing that I could think of would be "assume your data is a Gaussian distribution..."

enumaris

I should have at least said something like "I know the resources to answer this question" or something like that but nah, I just blurbed about gaussian distributions and standard dev...

then the dude cut me off cus the time was up

and that was that

bolbteppa

@Slereah cpht.polytechnique.fr/?q=en/node/163

enumaris

epic fail

danielunderwood

huh I don't think I've ever had a "time is up" interview

Slereah

@bolbteppa already went there!

danielunderwood

19:08

They always listen to what I say even if it's stupid

Slereah

Sent an email to the math phys one and the QFT one

Also sent an email to the string group just in case

Hire me dammit :V

enumaris

yeah I think it pretty much means he didn't need to hear more lol

Slereah

sing and dance

danielunderwood

On the bright side, what you consider a fail was my best attempt at a shot in the dark it seems

enumaris

o.o

danielunderwood

19:10

I have no idea how you'd do anomaly detection without first assuming a distribution

enumaris

you can look at the distribution of the data you have already

danielunderwood

But what if it's something scary?

enumaris

I was gonna go into a little bit of non-parametric methods

like anova

danielunderwood

come crazy n-modal distribution

enumaris

but

there are statistical methods to deal with those

I am not super well versed in them, but I know they exist

danielunderwood

19:12

Ahh right. we did anova in my stats class, but I don't remember any of it. I think it was just "import this library and call anova()" though

enumaris

generally under the heading of "non-parametric statistics"

danielunderwood

There is a statistics with R specialization on coursera that I may look at after this deep learning one

enumaris

I didn't get to that part tho

bolbteppa

@Slereah lpt.ens.fr/spip.php?article247&lang=fr e.g. the phys.ens.fr/~kashani/page3.html one, would look into it more than presented here

enumaris

ran out of time...

meh

danielunderwood

19:13

My view from physics is that only Gaussian distributions seem to really exist aside from that time we used a Poisson distribution

And I suppose we used binomial distributions a lot, but didn't really call them that

enumaris

my first interview went quite well, this one went quite poor...maybe they will average lol

Slereah

@bolbteppa oh god the page layout is so bad

bolbteppa

You have your in for the webpage-for-phd trade ;)

danielunderwood

Hey you haven't had a bad interview until you have one where you get off of the call and go "well that was the weirdest thing"

enumaris

I would say first one went 9/10, this one went 4/10 maybe they will give me 6.5/10 lol

danielunderwood

19:15

Happened to me when somehow the job requirements weren't correctly communicated to HR

bolbteppa

Would read up on the people here lpthe.jussieu.fr/spip/spip.php?article12&lang=en and send some off maybe too even if they're not in the huge list edpif.org/en/recrutement/prop.php#view39

Slereah

@bolbteppa I did try with Antoniadis already!

Might try with the others

enumaris

the ghost locomotive one was one where I went "well that was weird"...I would have rated that 3/10

bolbteppa

Yeah

Slereah

They don't have a PhD listed tho

enumaris

19:16

they requested a follow up interview tho so that was also weird

lool

danielunderwood

Ghost locomotive?

Also I had one that went really well, but ended up not getting the job

enumaris

self driving start up

yeah interviews are weird

my perception of what went right or wrong is pretty unreliable

I thought my microsoft interview went pretty well and they were like "nah, you ain't shit"

danielunderwood

Ahh I thought it was a typo lol

enumaris

:D

danielunderwood

I even had one that supposedly I was the only interviewee to answer all the questions well

Nick

19:19

hey, anyone here familiar with PSO vs ACO studies? I'm looking for physics applications.

Maybe I should ask the math people. since, it's optimization.

Slereah

@bolbteppa hasn't been my experience so far but it is worth a try!

bolbteppa

Hints

Nick

My project this sem is comparing PSO with GA, PA and any other ABCD I can find.

bolbteppa

You need to sell it right

Slereah

I wish it was acceptable to just send the same email to everyone really

Well I try to sell myself if I know the subject to some degree

But that's pretty hard except for a few instances

bolbteppa

19:22

i.e. read up, not just random/general ones, but don't bluff

Slereah

It is what I do, yes

Captain Bohemian

@Slereah would they find if you send the same mail to a bunch of professors? I guess they won't.

Slereah

But it's only good for selling it in a few instances

Like it was easier for Fewster since I actually read his stuff

bolbteppa

I wonder if you could get one at Lisi's Hawaii institute thing haha

danielunderwood

@enumaris do GANs compare with LSTM for sequence generation?

Slereah

19:24

@bolbteppa bit far off!

though really so far the problem is mostly just that either they have no thesis at all or they just ask me to submit a letter for funding

In only one case was I told I just plain old wasn't good enough

Captain Bohemian

I heard from Facebook a chemistry phd graduate sending emails to all professors in every chemistry faculty because he accepts any research topics, and finally got one accepting him who helped him apply for funding.

bolbteppa

@Slereah who said the good enough thing, the An.. one?

Slereah

@bolbteppa yeah

bolbteppa

I wouldn't read it that way at all

Slereah

The one who said I needed to have followed a string theory course during my master :p

enumaris

19:26

@danielunderwood GAN is not the same "level" of model as LSTM. It's a general framework for generating some data using an adversarial framework. An LSTM is a specific kind of cell in a RNN, it's not some general framework. You can have LSTMs in a GAN for example...

Slereah

And that books weren't good enough!

danielunderwood

Ahh I see

bolbteppa

Shame there's no online string coursera thingy :p

Slereah

Well I wouldn't be against doing a year of master to be up to speed

But it's like

I'd need 100% assurance that I'm getting a PhD behind it

Can't fuck around like that anymore

I've got bills to pay

danielunderwood

I suppose that should have been obvious now that I think about it in retrospect

bolbteppa

19:30

@Slereah part III?

enumaris

as you get more exposure, those things will come :D

Captain Bohemian

in Belgium there is postmaster program.

danielunderwood

@bolbteppa I'd really like to see some advanced courses on there, but I feel it doesn't have a huge demand. EdX has had some stuff, but it's all undergrad level from what I've seen

Slereah

@CaptainBohemian Same thing, tho

danielunderwood

There was a GR course on coursera, but I didn't find it to be particularly easy to understand

Slereah

19:30

Can't spend 6 months not working without an assurance of a PhD

enumaris

if bills are your concern...wouldn't working in industry be better financially than a PhD

I'm not sure your salary expectations really go up with a Physics PhD tbh...

Captain Bohemian

I even see prepostdoctoral program in some countries.

Slereah

I'm fine with a PhD salary

But I'm not fine with 0€ salary

bolbteppa

Probably best to forget the msc, all you want is one yes

Yashas

bolbteppa

19:32

(But part III is different)

Slereah

I can live in a shit flat for the duration of the PhD and eat ramen

that is fine

Yashas

is taking a definite integral on both sides of an inequality mathematically legal?

Slereah

but I can't do that on 0€

@Yashas Same rules apply as usual

danielunderwood

@enumaris yeah I need to just get to programming to get that exposure. I realize a bit of a confusion this morning that the dimensions of weights in an RNN is fixed. Doesn't that mean that the sequence length is fixed. So far that's been the case in the RNN lectures, but my understanding was that RNNs are supposed to work for variable inputs...maybe I just haven't gotten to that yet

bolbteppa

You could offer to do a free internship with those string guys if necessary as part of the process or something

Slereah

19:33

Depends if it's monotone increasing or something

@bolbteppa Bit risky

bolbteppa

Aim high, go for the Sorbonne, Bourbaki style

Slereah

They're not the ones deciding on the funding

bolbteppa

Get into the Cartan Bourbaki Goursat History

If you throw yourself at them (metaphorically) and your transcript is ok you could land somewhere

Slereah

I'm throwing myself a lot around

Yashas

@Slereah was that reply for me?

Slereah

19:36

Hopefully it will work

bolbteppa

Yeah it's a long game

Slereah

I'm not even sure I'll get one for 2019

bolbteppa

Amsterdam? iop.uva.nl/vacancies/strings/phd/phd.html

enumaris

@danielunderwood no, the sequence length can be variable. The sequence length does not enter into the dimensions of the weights. Weight dimensions depend on the internal hidden state dimension and the dimension of the input at one time step.

Slereah

I looked at Amsterdam yeah

enumaris

19:38

@danielunderwood what does need to be fixed is the input dimension from time-step to time-step. E.g. if you are inputting one-hot vectors encoding English words, your dictionary size can't be changing from time step to time step.

bolbteppa

Roughly there's like only a month maybe 2 left to get apps in then it's a wait for call backs (except for the ones that drip out from then on) so worth going all out

enumaris

@danielunderwood there is a caveat though that during training you might want to make sure your sequences are uniform in length since you want to do batching. This means you probably have to implement padding. Otherwise you would have to be reduced to pure SGD training which would be computationally very slow.

danielunderwood

Ahh I was mixing up the input dimension between timesteps with the entire input dimension

enumaris

yep

input usually (batch_size,time_steps,dim_per_time_step)

Slereah

@bolbteppa Well after that it's only funded ones, IIRC

enumaris

19:42

weights don't care about the first two in that tuple

Slereah

which is better to apply to but usually boring topics

danielunderwood

Do Keras and such automatically handle padding?

enumaris

not that I know

you need to input valid numpy arrays as your training set

padding has to be done a that stage since you can't have numpy arrays be jagged

there could be some util to do it that I haven't seen yet though

danielunderwood

So if you had something like one-hot encoded values, I suppose you'd add a padding token?

enumaris

you generally have a <UNK> token or a <END> token or something like that

keras embedding layer has that feature

it turns the [1,0,0,...,] vector into a "padding" variable I think

it's been a while since I worked on it though

Captain Bohemian

19:57

@Slereah do you think theoretical condensed matter physics boring topics to apply for?

Slereah

depends on the topic I guess

ACuriousMind

yeah, you don't wanna accidentally do something that's useful to someone :P

Slereah

Sorry society

I want to exclusively not be useful

ACuriousMind

I mean, do the wrong CMT, and you might invent a new semiconductor or something!

Slereah

I'll just have to work on toy models that don't exist at all

bolbteppa

19:59

The thing is, that stuff actually does influence qft and strings though it's hard

But you touch the dreaded 'application' as well

Slereah

But well

ACuriousMind

What do you mean "influence QFT"? Some CMT is straight QFT, just not HEP.

Captain Bohemian

but if getting a phd in theoretical condensed matter physics, it may not be possible to get a position in high energy physics after graduation.

Slereah

if I don't get anything by january

ACuriousMind

Beautiful acronym flood.

Slereah

20:00

I guess I'll start applying to more applied stuff

Condensed matter stuff is probably funded usually

Captain Bohemian

this April I saw a position in theoretical condensed matter physics in Amsterdam, but i didn't apply for it.

bolbteppa

Those lists above are full of things like that

Slereah

yeah I know

Yashas

does a qubit's state space have a particular name?

Captain Bohemian

I feel theoretical condensed matter physics is application of theoretical high energy physics.

Slereah

20:03

Condensed matter physics, experimental QM, nanophysics

those are the cool kids

@Yashas Bloch sphere?

bolbteppa

SSB came from condensed matter, big part of Landau's research was in it, it's nuts stuff

ACuriousMind

@Slereah That's the projective space, not the Hilbert space :P

Captain Bohemian

doing a phd in fundamental topics then one has more chance after graduation, I think.

Yashas

I need a Hilbert Space

Slereah

@ACuriousMind Universal cover of the Bloch sphere

then

Yashas

20:04

wikipedia says two-dimensional hilbert space but it isn't a name, is it?

Slereah

Or $\mathbb{C}^2$ I guess

ACuriousMind

@Slereah The n-sphere is its own universal cover for $n>1$.

Yashas

the most general one is $L^2$?

ACuriousMind

@Yashas Why does it need a name?

Slereah

I mean qubits are just $\mathbb{C}^{2n}$ hilbert spaces

They don't really need a specific name

It's common as dirt

Yashas

20:06

@ACuriousMind I was writing an answer where I was trying to show that a qubit's state space is a subspace of $L_2$ but writing "qubit's space" sounded awkward.

ACuriousMind

@Yashas $L^2$ is not a complete notation (you want e.g. $L^2(\mathbb{R})$), and those spaces of square-integrable functions having nothing to do with qubits.

@Yashas The qubit state space is not a natural subspace of any $L^2$ space (though of course you're free to map it to any two-dimensional subspace of any $L^2$ space).

Yashas

@ACuriousMind the qubit states must also satisfy the constraints such as the state functions must be normalized?

ACuriousMind

@Yashas There are no functions. There's just a state vector. The qubit state space is finite-dimensional and not expressible in terms of wavefunctions.

Yashas

I appear to be lost once again :O

ACuriousMind

All these function spaces are infinite-dimensional vector spaces. The space of a qubit is not such a space.

Yashas

20:08

where does a|0> + b|1> requiring a^2 + b^2 come from then?

Slereah

In quantum computing you just ignore all degrees of freedom of the Hilbert space except spin

so it's just a 2D Hilbert space

ACuriousMind

@Yashas Well, that's normalized. But it's not a function.

Yashas

I thought I could relate all of qubit stuff to the more general QM stuff.

Slereah

Well you can, but general QM doesn't need $L^2$ spaces

Many different spaces are Hilbert spaces

ACuriousMind

@Yashas The general QM conception is that the state space is a Hilbert space, i.e. states are vectors. In many cases, this space is infinite-dimensional and can be represented as a space of functions. In some cases, in particular those in quantum information, they are finite-dimensional and can not be represented as spaces of functions.

It is in fact the functions which are the special concept here, not the vectors.

Slereah

20:10

Hell you can even do a 1-dimensional Hilbert space if you want

It's boring but you can

$\mathbb{C}$, with states on $S$ and projective Hilbert space $\{ 1 \}$

And it only has one operator, $I$

Not much you can do with it

Yashas

@ACuriousMind Hilbert space is the most general one?

does it mean that I can describe QM using any Hilbert space?

ACuriousMind

@Yashas For some values of most general...

You can do a formulation in terms of $C^\ast$-algebras where you don't have to fix a Hilbert space, but you very likely do not need to worry about that

@Yashas I don't understand the question

Slereah

The most general Hilbert space in QM is just an abstract space that obeys the properties of a Hilbert space

(Well, that and it's separable)

ACuriousMind

QM is a framework, like classical Hamiltonian mechanics

There are many different physical systems, each of which has its own state space (a Hilbert space in QM, a symplectic space in Hamiltonian mechanics)

Slereah

Or even worse

a MULTISYMPLECTIC SPACE

shudders

ACuriousMind

20:15

In principle, you can think about any Hilbert space together with a designated Hamiltonian (some Hermitian operator) as a QM system. It is of course not guaranteed that there is any real-world system modeled by that particular pair.

Slereah

I mean technically there's only one Hamiltonian corresponding to a real world system

And you have to apply it to the wavefunction of the universe

ACuriousMind

@Slereah No there isn't.

Slereah

But it's a bit tricky

ACuriousMind

Hamiltonians, like Lagrangians, are not unique.

Slereah

Oh sure, UP TO SOME GAUGE

Or whatever

mister picky

ACuriousMind

20:20

You said technically

How am I supposed not to nitpick that? :P

Slereah

Hisss

You win this round, Dick Dastardly

danielunderwood

I don't know what a multisymplectic space is, but it sounds scary. I'm not even that familiar with plain old symplectic geometry

ACuriousMind

::twirls mustache:: Until we meet again!

Slereah

Multisymplectic space is the one you use for field theories

Where the phase space is field configurations + configurations of the field momentums

it's pretty bad if you look into it too close

solvayinstitutes.be/html/doctoral.html

Hm

Doctoral school is about 3 months

Still pretty long to leave work tho

Unless I put like ALL THE VACATIONS THERE

Not ideal

why is everything so hard

ACuriousMind

...you get 3 months vacation?

bolbteppa

20:23

Need to be in a phd already I think

Slereah

@ACuriousMind Well I have about 6 weeks vacations

Total doctoral school is 9 weeks

ACuriousMind

6 weeks != 3 months, technically

2

Slereah

So not really

Unless I take like 3 weeks without pay :p

Though if I'm ever between jobs that would be ideal

Emilio Pisanty

21:32

bam

nine zero rep caps, baby

Slereah

badaboom badabing

Emilio Pisanty

That's pretty close to 150, no?

enumaris

welp, meetings done for the day

fun beans...

danielunderwood

Just got done with my interview...seemingly went well

Though there were a variety of technical questions but the next round is supposedly the technical round

enumaris

nice nice

enumaris

21:51

seems like the infosec team is not gonna go forward with my suggestion to not change passwords every 3 months...

also probably no free wifi...sad

:(

danielunderwood

Actually the job I just interviewed for was a security position

Maybe if I had a position like that then went into data, I could just point at it and be like "I know security" :D

enumaris

lol

are you interested in infosec?

danielunderwood

Yeah it's another area I find interesting...though more in the form of break things and figure out who let me do that than tell people what they can and can't do

enumaris

my exposure to infosec is quite boring lol

Sha Vuklia

guys, I'm not really sure where to find more information, but I'm trying to figure out what is the commutator of two vector operators. So say we have $A=[A_1\, A_2\, A_3]$, where each $A_i$ is an operator, then I would think that $[A,A]=0$, because $[A_i,A_i]=0$. But apparently we have $[A,A]_{i,j}=[A_i,A_j]$, so we have to consider all components. I found this last expression as the definition of $[A,A]$ on a stack post, but I don't really understand why my own thinking doesn't hold.

I used to think that a vector operator is just a vector whose components are operators, but I'm finding online that there are a couple more conditions for something to be called an operator (but it's still quite confusing to me). Could someone clarify, or maybe redirect me to a readable source?

danielunderwood

22:02

The position I applied for was actually a data role in security, but the interviewer suggested a different role may be a better fit in a way that suggested that I could either be a top candidate for that role or a not top candidate for the role I originally applied for

enumaris

my initial interview with the AI consultancy had the interviewer say they think I'm a better fit for a slightly different role

and they asked for a 2nd interview

so

danielunderwood

Yeah that's basically what happened

enumaris

welp, can't change anything now

just gotta wait and see

we are in the same boat :D

danielunderwood

But the suggested role sounds pretty good as well and it sounds like there's some flexibility in roles

ACuriousMind

@ShaVuklia The commutator of two "vector operators" is not actually a standard concept because you'd have to say what it means to actually apply the full vector to a state twice, and it is not clear what $\vec A \vec A \lvert \psi\rangle$ is supposed to be (the definition from the stack post you cite implies it's a matrix of states, but it's not clear that that's the "natural" notion). You can define it however you like.

And a vector operator is just a vector whose components are vectors. Who's trying to tell you otherwise?

enumaris

22:07

there's generally some flexibility, unless the company is hiring for a specific project or something

danielunderwood

Well it sounded like a specific project, but the project is something that's just getting started it seems

ACuriousMind

@enumaris I choose to believe that an AI consultancy is a consultancy founded and led by an AI.

enumaris

I see

@ACuriousMind it could be the case...I have not met the founder yet

ACuriousMind

dun dun dun...

enumaris

I've met (conference called) with 3 data scientists and 1 principal data scientist

danielunderwood

22:09

Or 1 good machine

enumaris

could be

1 machine that passed the turing test 4 times

ACuriousMind

Clearly, "data scientist" is just a multivalued function.

(That's a pun on prinicpal value)

enumaris

(explaining jokes takes the fun out of it bro)

should let me stew

wonder what exactly you mean

and maybe in 15 years I'll let out a chuckle

ACuriousMind

:47792851 1. As I said, there's no standard notion, so it is what ever you'd like it to be. Applying a vector of operators to a state is rarely useful. 2. Is the $i$ in your expression summed over? If yes, then one would more expect that to be written $(\vec A\cdot \vec A)\lvert \psi\rangle$.

danielunderwood

I wonder what gradient descent does with a multivalued function

enumaris

22:12

it just does it element by element

Sha Vuklia

@ACuriousMind yes, I just realised your second point. If we apply $\vec A$ once, then it's quite straightforward what we get, because $\vert\psi\rangle$ is scalar, but if we apply $\vec A$ twice, then on the second application we would have to define our product. that's how I'm seeing it now

enumaris

you get a Jacobian matrix rather than a Jacobian vector

danielunderwood

Wait aren't all Jacobians matrices?

ACuriousMind

@enumaris I think you're confusing a multi-valued function with a vector-valued function...

enumaris

$(\delta J)_{ij} = \partial J_i/\partial W_j$

ACuriousMind

22:13

Also, what @danielunderwood said.

enumaris

right, so if you have a vector-valued function you have a Jacobian matrix

ACuriousMind

@ShaVuklia I think the more important question here is why you'd want to define that at all.

enumaris

if you have a scalar-valued function, your "jacobian matrix" is really just a row vector...

Sha Vuklia

@ACuriousMind I just came across this, because my reader asked me to show that $L$ doesn't commute with itself

ACuriousMind

@enumaris It is still, semantically, a matrix, since it operates on tangent vectors.

danielunderwood

22:15

Oh you're calling the gradient a Jacobian. And I'm not really saying a vector-valued function. I'm saying something like an inverse trig function or square root. Though I suppose you could use the branches as components or something

Sha Vuklia

I guess it's not that useful, but I wondered because of the exercise

enumaris

ah

ACuriousMind

@ShaVuklia Ugh. I hate such ill-defined questions :P

Why wouldn't they just say: "Show that $[L_i,L_j] \neq 0$ for $i\neq j$."

enumaris

@danielunderwood I've not dealt with complex loss functions so I'm not sure

Sha Vuklia

I think I know what they were on about

let me show

ACuriousMind

22:16

Because that's certainly what they intended you to do.

danielunderwood

Though the square root would be particularly odd since it would "fall off" if you didn't look at the imaginary component too

Sha Vuklia

no no, they had sth else in mind:p slightly different

let me show;

first they want me to write $L$ as a cross product of itself

and then I can show that $[L,L]\neq 0$

enumaris

what I was referring to is

Sha Vuklia

but I was thinking, screw that cross product, I just want the definition of [L,L] :p which apparently turns out to be $[L,L]_{i,j} = [L_i,L_j]$:d

enumaris

if you use like some generic loss in tensorflow but forget to average it over some dimension, tensorflow will still work iirc...

ACuriousMind

22:18

If you did not explicitly define the commutator of vector operators somewhere, I still maintain that's an ill-defined exercise :P

enumaris

tbh it has been a while since I've done something like that on purpose

danielunderwood

I suppose if you change your choice of principle value, you'd end up with something like this datascience.stackexchange.com/questions/23257/…

wrong link

Sha Vuklia

hahah I agree, it's ill-defined, but they strongly suggest a certain approach:p

danielunderwood

Multi-objective optimization

Multi-objective optimization (also known as multi-objective programming, vector optimization, multicriteria optimization, multiattribute optimization or Pareto optimization) is an area of multiple criteria decision making, that is concerned with mathematical optimization problems involving more than one objective function to be optimized simultaneously. Multi-objective optimization has been applied in many fields of science, including engineering, economics and logistics where optimal decisions need to be taken in the presence of trade-offs between two or more conflicting objectives. Minimizing...

I think I have a problem in being interested in edge/pathological cases

ACuriousMind

@danielunderwood Uh...just do constrained extremization with Lagrange multipliers?

enumaris

22:21

yeah I dunno what the application might be, but I just recall tensorflow didn't error out when I forgot to average over some dimension I think

so my last post (2 posts back) should actually say I can't recall ever doing something like that on purpose

it was all accidental

danielunderwood

hmmm I'm not all that familiar with Lagrange multipliers

I had a reason to learn them for something, then promptly forgot

ACuriousMind

Ah, nevermind, I misunderstood the problem setup

enumaris

lagrange multipliers allow you to optimize a function given some constraint

the multipliers implement the constraints

danielunderwood

Yeah but I don't remember how to do the actual calculation. It's fairly simple iirc though

Emilio Pisanty

39

Q: Time for some more swag!

On occasion of our anniversary celebrations, we offered some cheese boards a while back. If you were late for that, though, no sweat! You’re just in time for another limited edition super special swag contest! This is the bit where you go “It’s about time, isn’t it?” To which I’ll reply with “we...

oh, man

that is so tempting

danielunderwood

22:29

PSE sure does have fewer questions tagged time than I would have thought

Emilio Pisanty

I'm thinking a dedicated Q&A on that tag would make a nice entry

ACuriousMind

> number of violations of PEP8 are embarrassing

Every Python programmer, ever.

(@Chair)

Emilio Pisanty

no idea if I have the time to write it before FGITW takes over though

I object - that's not what Big Ben looks like. This is what it looks like (as of this year, and until ~2021, it seems). — E.P. 10 mins ago

enumaris

does any python programmer actually read the PEP guidelines?

I read specific cases...

danielunderwood

I feed my code through pylint or similar

ACuriousMind

22:32

meta.stackexchange.com/a/318926/263383 ...they nuke spam on Stack Exchange using a smartwatch? Oh brave new world we live in.

I feel the deletion should be triggered by a punch-y gesture, though

danielunderwood

I don't know that I've ever looked at PEP8 much, but I do look at PEPs occasionally

also I liked the inclusion of this in one of the answers

Space Otter

If you plot a graph of intensity vs angle after a double slit experiment what is that type of curve called?

Like a sine waveform but which damps in both directions

I'll try find a picture

ACuriousMind

@SpaceOtter The enveloping function (responsible for the damping) is the square of a sinc

Also, the sine wave is actually a squared sine.

Slereah

I bet the guy who came up with the name cardinal sine had a good laugh

Space Otter

Hmm, It seems origin doesn't do that

Thanks though @ACuriousMind

ACuriousMind

22:42

@Slereah ...before he went to HELL

Semiclassical

helle

enumaris

uhhh

danielunderwood

helle's belles?

Uh oh...just got a "take home assignment" from that interview

Transcript for