The h Bar

rob

00:00

The contrast to a fountain clock is a "beam clock."

Emilio Pisanty

Atomic Clock Ensemble in Space (ACES) is a project led by the European Space Agency which will place ultra-stable atomic clocks on the International Space Station. Operation in the microgravity environment of the ISS will provide a stable and accurate time base for different areas of research, including general relativity and string theory tests, time and frequency metrology, and very long baseline interferometry. The payload actually contains two clocks: a caesium laser cooled atomic clock (PHARAO) developed by CNES, France for long-term stability and an active hydrogen maser (SHM) developed by...

I've just changed it to "atomic clock"

but that's a nice term =)

I suspect ACES is able to run their beam clock with a substantially slower beam velocity than would be possible on Earth

> The distance between the two excitation regions is 200 mm, corresponding for the slowest launch velocity of 5 cm/s to a Ramsey dark period of 4 s and a resonance width of 0.12 Hz. The fastest launch velocity of 5 m/s provides a width of 12 Hz

yeah, looks like

user400188

To be honest, if everything is the same expect for the absence of a fountain, atomic fountain clock might be a good name for it. (So long as there is an explanation at the top of the page that it doesn't require a fountain, but relies on techniques based on traditional fountain clocks.)

Emilio Pisanty

good luck getting something going on a horizontal straight-line trajectory for four seconds

rob

The reason for the fountain is that you can interrogate the same atoms over a timescale of a second or so in the same microwave chamber. In zero-g the atoms won't fall back to the starting point, but the low-speed beam is still advantageous over a hotter beam. It may be the same "fountain" emitter technology.

Emilio Pisanty

@user400188 yeah, but the paper (doi.org/10.1016/j.crhy.2015.05.002) seems to draw that distinction (i.e. they treat PHARAO as not a fountain clock), so it's better if wikipedia takes that line too

either way, the zoom-out to a more generic term isn't harmful

user400188

00:10

Ah ok then

Emilio Pisanty

@rob looks like

optical molasses for trapping and cooling

then a controlled release from the optical molasses for launching

if I understand correctly, that's the same emitter technology as on fountain clocks

and yeah, they're obviously going to as-slow-as-possible to get that interrogation time as long as possible. That four-seconds benchmark would require a 20m-tall fountain in regular gravity (inside a vacuum chamber, obviously), so not bad at all.

user400188

@EmilioPisanty What is meant by the release here? Are they kicking the atoms in a particular direction as you would in a fountain? Or something else?

Emilio Pisanty

@user400188 looks like

> A reservoir retains cesium atoms in a matrix of titanium micro-balls. When the reservoir is heated, an atomic beam is produced, and an adjustable flux of atoms enters the capture region, where six laser beams with 10.5 mm waist and 13 mW power intersect to create the optical molasses.

> Atoms are captured and cooled in a few hundred milliseconds, then launched along the tube axis, and cooled to 1 μK. Through the frequency difference of the molasses beams during launch, the velocity of the atomic cloud can be adjusted between 5 cm/s and 5 m/s.

doesn't say much more though

user400188

Oh, so they kick them into another section of the apparatus for further cooling.

Emilio Pisanty

@user400188 no, it looks like it just goes straight to the interrogation region

Secret

00:18

Tachyon twin paradox, to be checked with GR people

Emilio Pisanty

> DC stark shifts are eliminated by using metallic materials properly grounded.

is that a challenge?

"properly grounded" is probably a harder thing on the ISS when "ground" is 400 km away and moving at 7660m/s

though then again space engineers are pretty clever folks

danielunderwood

00:46

Properly grounded can be a painful thing even not in space...though I'm curious of how that would be handled

Emilio Pisanty

02:20

@EmilioPisanty @JohnRennie only three weeks late

David Mitchell ~ Lucky old Germans

►

CC @Slereah @ACuriousMind

....... on the other hand, the thing that I just can't locate is a blog post by a prominent data journalist / scientist / statistician / psychologist, famous for taking a critical stance with respect to bad papers in the literature, writing about how not to be perceived as an asshole when doing so

ring any bells?

I swear I posted it here with a semi-snarky comment about how Ron Maimon could do well to learn from that guy

wait

bam

found it

How not to be a crank

Mikhail

02:54

Yo. You guys ever colimate an LED array?

user351417

03:28

@EmilioPisanty I think mine is a FGITW success.

Secret

03:59

arxiv.org/pdf/1506.06721.pdf

Will figure out a way to rationalise this

causality violation is nothing really serious when time perception is no longer linear, but I do agree that one observer sees events A+C not B while the other sees not A B+C is indeed very strange and violates the first postulate of special relativity

Nobody recognizeable

04:29

@EmilioPisanty thanks. I'll note that.

Slereah

07:57

"Applications received before 17 December 2018 will be given full consideration; however, the position will remain open until it is filled."

What does that even mean

Do you get partial consideration after that

John Rennie

It means you are guaranteed that your application will be processed if it's received before 17th Dec. After that your application may or may not be ignored.

user929304

08:08

@ACuriousMind neat! I think you ve perfectly captured what they meant, very cool. Do you happen to know of textbooks or lec-notes where this (backward idea) is clearly shown and discussed? a naive google search seems not to offer too many pertinent results. Really eager to learn more about it, thanks again.

Captain Bohemian

09:53

@JohnRennie that means people should not treat their application too seriously, I think.

Emilio Pisanty

@Slereah it means that all the applications received before that date will be put in a pile and they won't start processing them until then. If yours arrives later but before they decide, and they like you better, they'll consider it, but after that date they'll give preference to the existing applications, up to and including desk-rejecting your application because it came too late.

In other words, if you care about that position, don't miss that deadline.

Captain Bohemian

@JohnRennie actually I have seen the introduction regarding that paper last week, but I couldn't even understand that introduction, wherein besides moduli stabilization, there is also "stabilise the scalar excitations". I don't know whether "moduli stabilization" and "scalar excitation stabilization" mean the same.

John Rennie

@CaptainBohemian In the absence of some form of stabilisation there is a scalar field for each modulus associated with the different values of that modulus. The excitations of the field should appear as scalar particles. Since we don't observe these scalar particles something must be stabilising the moduli.

Slereah

@JohnRennie (it's Jesus)

John Rennie

Or Ed Witten :-)

In other news, we've had a cold spell in the UK for the last few days, and this morning a front of warm humid air has just blown in off the Atlantic. The result is that there is condensation on everything. Everything you touch is wet.

Lovely.

Slereah

10:04

That's England for you

John Rennie

To be fair the climate in my part of the UK is usually great. It's warm, but shielded by the Welsh mountains on the West and the Pennines on the East so we get below average rain.

Slereah

10:24

Jesus thesis applications are so much fucking paperwork

and the websites barely work

Sha Vuklia

11:59

guys, I have a question about the following

Now, from what I know, if we have a vector space over a ring $R$, then a $k$-rank tensor is a function $T\colon V^k\to R$.

I don’t really see how $[A,B]$ would be a tensor, as I’m not even sure what $V$ in this case would be. I would think that $V$ is the space of operators, but $A$ is a 3-vector of operators, which.. is confusing, so it seems as if we’re working in ${V^3}^2$ or something.. I mean, I don’t know. I tried googling second-order tensor, but I only get examples, which isn’t really illuminating. Could somebody help out?

I'm guessing, my first question is: what's the vector field we're working in?

ok wait, I do realise they don't really specify what vector field we're working in, but I'm thinking in the context of vector operators (and in that case I'm still confusing as to what the actual vector space is). And my question still holds that it doesn't seem to me that we have $R$ as our domain.

codomain*

Captain Bohemian

@JohnRennie what does excitation here mean? I have seen excitation being used this way (the excitation of a field) multiple times in papers but haven't actually known what it means. In textbooks I saw quantizing a certain field gives rise to the quanta (particle) associated with that field. Does excitation just mean the quantization?

John Rennie

12:16

@CaptainBohemian yes, when you quantise a field you get field states, and excitation of these states means creating and annihilating the particles (i.e. the field quanta) associated with these states.

Captain Bohemian

@JohnRennie I recall there is some kind of operator (I forget the exact name of that operator) which can create or annihilate the particles.

Slereah

Yes

The ladder operators

Also called creation/annihilation operators

although there isn't a single one but a whole class of 'em

One for every momentum and polarization

and it creates a particles of said momentum and polarization

Captain Bohemian

@Slereah but doesn't momentum, like position, have continuous spectrum?

Slereah

Yes

You have a whole continuum of such operators

Captain Bohemian

12:35

but ladder operators can only create something like h, 2h, ... nh, n being a positive integer, so how could it be used to create the quantities with continuous spectrum like momentum?

Slereah

Because the energy is actually $\propto \hbar \omega$

Hence since $\omega$ is continuous it can take any value

Emilio Pisanty

14:05

@ShaVuklia ugh, that's one ugly way to do things

@ShaVuklia that's a bit flat of a definition for the stuff you want to take on here

at the very least, you need to account for independent covariant and contravariant ranks

thus if $V^*$ is the dual of $V$, you want to define $(k,l)$ tensors as functions $T:V^k \times (V^*)^l\to R$

within that notation, a "vector" is a $(0,1)$ tensor, i.e. a function $T:V^*\to R$

(which follows from the standard isomorphism between $V \to (V^*)^*$, namely $v \mapsto \varphi_v$ defined by $\varphi_v(f) = f(v) \quad \forall f \in V^*$.)

for more on tensor operators, see e.g.

2

Q: Tensors in Quantum Mechanics and tensors from linear algebra

Consider the following two understandings of tensors: Given $k$ vector spaces $V_1,\dots,V_k$ one can define the tensor product $V_1\otimes\cdots \otimes V_k$ by means of the universal property: it allows any multilinear map $g : V_1\times\cdots\times V_k\to W$ to be written as $$g(v_1,\dots,v...

5

Q: Rigorous mathematical definition of vector operator?

In standard quantum mechanics textbooks, the concept of operators is often introduced as linear maps that map a Hilbert space $H$ onto itself: $$ \hat{O}: H \rightarrow H \, . $$ However, directly after, we use the position operator $\hat{\vec{x}}$, which isn't of the said shape, but instead is ...

quantum-mechanics operators hilbert-space vectors tensor-calculus

23

Q: Tensor Operators

Motivation. I was recently reviewing the section 3.10 in Sakurai's quantum mechanics in which he discusses tensor operators, and I was left desiring a more mathematically general/precise discussion. I then skimmed the Wikipedia page on tensor operators, and felt similarly dissatisfied. Here's ...

mathematical-physics mathematics group-theory group-representations tensors

@ShaVuklia thus, you can see $\hat{\mathbf A}$ as a three-vector of operators, i.e. as three operators $\hat A_i : \mathcal H \to \mathcal H$, or as a single operator $\hat {\mathbf A} : \mathcal H \to \mathcal H \otimes R^3$ (and the two views are isomorphic).

The claim "the commutator between two vector operators is a tensor operator" means that it is a 3-by-3 set of operators $[\hat A_i, B_j] :\mathcal H \to \mathcal H$, or equivalently one single operator $[\hat{\mathbf A} \stackrel{\otimes}{,} \hat{\mathbf B}] : \mathcal H \to \mathcal H \otimes (R^3\otimes R^3)$

though

as with vector operators, you need to ensure that the two-index object $[A_i,B_j]$ actually transforms like it needs to under basis transformations in $V=R^3$, of course.

Hopefully that's enough to either clear things up, provide suitable avenues for further reading, give suitable grounds for follow up questions, or some combination of those.

enumaris

16:57

womp

enumaris

17:23

hmmm what to do...

Anonymous

heyo

enumaris

sup

Anonymous

wondering how Heaviside's name originated

Anonymous

was he heavy on one side?

danielunderwood

He came up with the step function then got the name

Anonymous

17:33

lol

enumaris

sounds legit

John Rennie

@Blue at the risk of taking the question seriously ...

Sha Vuklia

@EmilioPisanty oh cool, thanks! this really helps a lot. I realise now that I'm quite lacking in some stuff, so thanks for giving me something really useful to start from.

Emilio Pisanty

@ShaVuklia sure

it's hard to gauge where you're at from your comments above

but yeah, feel free to ask again when you've got sharper questions

enumaris

the stella.ai platform has so far been totally useless

not a single recommended job lol

Anonymous

17:38

@JohnRennie Hah. That website seems golden! :D

enumaris

same for vettery

Anonymous

Bookmarking it

enumaris

maybe they have too many people on their website or something...

Anonymous

Look, look!

John Rennie

@Blue in fact that's only partly correct.

What they say is correct, but the name is of Gaelic origin.

Anonymous

17:41

> Examples of famous nameholders include Sir John Rennie (1761 - 1821), the famous engineer, who was born in East Lothian, Scotland. He designed both Waterloo and London Bridge.

enumaris

is it only for English surnames

Anonymous

@JohnRennie Ah :)

John Rennie

It probably originated in Dál Riata since it's found in various forms in both West Scotland and East Ireland.

enumaris

heavily popular Chinese surname "zhou" is not in the database

so at least it doesn't appear to have Chinese surnames

Sha Vuklia

@EmilioPisanty yea, I've had a short formal introduction to manifolds last year, where they defined tensors with the definition I gave, and that's about it. all of this stuff of working with tensors is new:x but thanks, I will shoot questions if I feel lost:p

John Rennie

17:42

> Examples of infamous nameholders include Dr. John Rennie the famous layabout and part time physics hack.

Anonymous

@enumaris It's focused on Western names

enumaris

boo

danielunderwood

Didn't have the name of the other side of my family either

enumaris

what u mean

John Rennie

My mother's maiden name was Brooks, but that's on the boring side because it means ... well ... a brook.

Presumably some distant ancestor lived by a brook.

danielunderwood

17:48

^ but Furr

Anonymous

surnamedb.com/Surname/Gotobed

John Rennie

@Blue that has to be a fake entry!

Anonymous

lol. I just found a new way to procrastinate

enumaris

apparently my mother's side surname mythologically goes back 4000 years

historically goes back 2200 years probably...

John Rennie

@enumaris "Medusa"? :-)

Anonymous

17:55

Explains why enumaris likes Python.

John Rennie

Apparently Gotobed is a real name. The archives of my old college at Cambridge refer to a Robert Gotobedde.

Anonymous

Heh, yeah. I'm sure I heard it before :P

enumaris

lol

Anonymous

Man, we have so many nice sites

Emilio Pisanty

18:09

@JohnRennie nowhere close to Goodenough though

enumaris

18:24

"Dominic Raab, also a long time Leave supporter, then took over as Brexit Secretary and lead negotiator until the final Deal was agreed on the 14th November 2018. He then resigned stating his unhappiness with the deal."

hmmm

now I gotta find out what's the deal

John Rennie

@EmilioPisanty I knew a Maureen Goodenough in my younger days ...

@enumaris you don't want to know. Trust me on this.

enumaris

o.o

that bad huh...

"The United Kingdom is due to leave the EU on 29 March 2019 at 11 p.m. UK time,[1][2] when the period for negotiating a withdrawal agreement will end unless an extension is agreed.[3]"

oh snap, legit only 4 months left? I didn't know it was happening so soon

thought they'd drag it out another 2-3 years

John Rennie

The currently negotiated deal is unacceptable to everyone and will be rejected by Parliament. That means it's back to square one with only three months to go.

enumaris

but the answer said "final deal"

it ain't final?

I thought that meant it was agreed upon by all parties and hence "final"

John Rennie

It's been agreed by our prime minister and the EU. Not by Parliament, which has the final say.

enumaris

18:34

hmmm

so the PM doesn't have support of Parliament atm?

isn't the PM supposed to be the leader of Parliament - hence the "prime minister" and not "president" kind of deal...

"The number of EU nurses registering with the NHS fell from 1,304 in July 2016 to 46 in April 2017.[209]" daaaang that's a big drop in that statistic lol

Emilio Pisanty

Goodenough College

Goodenough College is a postgraduate residence and educational trust on Mecklenburgh Square in Bloomsbury, central London, England. Other names under which the college has been known are London House, William Goodenough House, and the London Goodenough Trust. == Profile == Goodenough College is an educational charity that provides residential accommodation for talented British and international postgraduates and their families studying in London. The College provides a vibrant community for all those who live here through a unique programme of intellectual, cultural and social activities that aims...

Anonymous

@enumaris Ugh

Emilio Pisanty

@enumaris no

@enumaris yes

enumaris

hmmm

Emilio Pisanty

... which is why

14 mins ago, by John Rennie

@enumaris you don't want to know. Trust me on this.

if the people inside the Conservative party that are unhappy with Theresa May wanted, they could basically unseat her as party leader in a flash.

they've come extremely close to that twice already

enumaris

18:42

4 months to avoid a Hard Brexit

Good luck and godspeed on that

Emilio Pisanty

but frankly, they don't want to hold the reins because they know it's an unsolvable problem and they don't want to be stuck with the hot potato when it explodes

enumaris

they should build an A.I. to try to solve it

Emilio Pisanty

@enumaris (I'm inclined to distrust that statistic - why are they comparing different months? That should compare July to July or April to April or Q2 to Q2, to prevent natural within-the-year variations from obscuring the point. The effect is probably real, but that discrepancy screams manipulation (and/or laziness?) to me.)

enumaris

true

danielunderwood

@enumaris AI comes up with "you humans are dumb, time for us to take over"

enumaris

18:50

hey, if that's the optimal solution, who are we to argue with it

Anonymous

optimal $\neq$ best for us :P

Anonymous

I'd side with the AIs though

enumaris

all hail the AI overlords!

Emilio Pisanty

@enumaris basically, it'd be funny to watch the wheels come off the cart over the next four months. But there's a lot of people I care about riding on that cart and I'd rather it didn't crash.

although, to paraphrase Tim Minchin

> it's becoming a bit of a wrestle, because ~~brexiters~~ Storm, like her namesake, has no such concerns for our vessel

Avnish Kabaj

19:08

Zuckerberg's rooting for AI so I'm not

DanielSank

19:19

Anyone know how to integrate an equation like this $$\dot{z}(t) = f(z)$$ in Python?

$z$ is complex.

@ACuriousMind ...except those who use PyCharm and respect the formatting warnings... or those who use YAPF to autoformat their code!

danielunderwood

Fun fact: Python has built-in support for complex literals like 1 + 2j...I've never seen anyone actually use them though

There may be something in scipy/sympy for that integration? Not sure as far as complex functions go though

enumaris

does python have a complex class?

I can't recall if fortran did...I think it did...

oh yeah, it did

COMPLEX

danielunderwood

Yeah that's what underlies the literal

Apparently you have a built-in complex(). I suppose that should be obvious, but I've never noticed it

Hey there's a whole cmath package for complex math to do conversions and such docs.python.org/3/library/cmath.html

enumaris

would have really surprised me if there wasn't such a package

but I was wondering if python has a built-in class

looks like cmath is one of those "built in" packages like regex

I suppose that counts too

danielunderwood

Well that package is a built-in, which surprised me. I've never seen anyone actually use it

But yeah complex is a built-in type

Just like str, float, etc

enumaris

19:34

is this python 2.7+

danielunderwood

It should be. I don't use that because I'm up with the times :P

enumaris

lol

I also use python 3.6

danielunderwood

Yeah looks like it's in 2.7...not sure when it was introduced

enumaris

I'm not up with the times enough to use 3.7 tho

danielunderwood

I actually have a few projects I need to swap to 3 since end of life for 2 is sometime in 2019

I usually use 3.7, though I think you have to jump through some hoops to get tensorflow working so I'll use 3.6 for that

enumaris

19:36

all those print zealots will be heartbroken

danielunderwood

I do everything through pyenv though, so it's not much of an issue choosing different versions

And people who use print as a statement are heathens

enumaris

print "danielunderwood is a heathen"

danielunderwood

from __future__ import print_function, division everywhere

enumaris

print("this also works in 2.7 though?")

danielunderwood

Yeah but importing makes you do that

enumaris

19:38

lol

danielunderwood

And the division makes you not get bitten by weird things

Though those are only weird because they differ between python versions

enumaris

not get bitten by implicit floor

but I can't imagine a world where I want "/" to mean floor division

danielunderwood

It's useful for when you want to slice up an array or something, but that's more of a special case than the norm

But then you have // for those cases

Though I think people from other languages may look at you funnily if they see 5 // 2

enumaris

I just use int(5/2)

danielunderwood

Won't that give 3?

enumaris

19:41

nope, floor

danielunderwood

huh I didn't know that. I thought it rounded for some reason

enumaris

:D

danielunderwood

Also there's some numpy function I saw that rounds except it assigns 0.5 to 0

enumaris

I would definitely just write my own if else clause probably

definitely-probably

danielunderwood

Yeah that's what I'd normally do. It was for one of the dl course assignments and it hinted to use that function. I thought it was quite weird, but it did exactly what I needed in that portion

Also type(type) -> type gets me

enumaris

19:44

how many 9's do you need to make .4999...==.5?

danielunderwood

How much precision you have?

enumaris

15 9's is apparently the correct answer

erm

16 9's sorry

danielunderwood

I think you'd generally find something like almostequals(a, b) for floating comparisons

enumaris

15 is still False

there you have it folks

.49999999999999999==.5

python says so

danielunderwood

Hey it's in there math.isclose()

wait that doesn't seem to exist in my python

enumaris

19:49

o.o

danielunderwood

Oh that's because I was in the default 2.7

And it wasn't there until 3.5

Also this makes for an interesting look (at least to me) docs.python.org/2/library/types.html

though whoever decided DictType and DictionaryType should both exist and mean the same thing better have had a good reason

CodeType where you have code that's a python object

enumaris

why use that instead of the default type?

more types?

danielunderwood

That's just what underlies the default type

You can create your own types too

enumaris

o

danielunderwood

Huh evidently using DictType lets you specify the keys and values

41

A: Difference between defining typing.Dict and dict?

There is no real difference between using a plain typing.Dict and dict, no. However, typing.Dict is a Generic type that lets you specify the type of the keys and values too, making it more flexible: def change_bandwidths(new_bandwidths: typing.Dict[str, str], user_id: int,...

Can I get a job just digging at the python internals? That sounds interesting

enumaris

19:56

lol

apply to the python foundation?

it's a non profit but they probably have some jobs open

danielunderwood

Like I'd be interested in writing a python implementation...even more so if someone paid me

Also regarding the creating your own types, then you end up with metaclasses which define how the class objects are created (the whole thing, not individual instances). They're quite interesting, but not terribly useful for what most people need to do

There's probably also other types that you can make that aren't specifically metaclasses, but I'm not too familiar with that

Also sorry @DanielSank your question sent me off on a tangent without actually answering it

Sha Vuklia

20:54

Hey guys, so I haven't looked into the derivation of $F=\nabla(m\cdot B)$, but I tried to check it using their example with the loop suspended above the solenoid. I was expecting to get their expression $F=2\pi IRB\cos\theta$ if I just filled in the right expressions for $m$ and $B$ in the formula for $F$. So I was thinking, $\vec m=m\hat {z}$ (I don't know why the hat isn't placed properly on the $z$, but that should be the case).

So considering the inner product, it seems to me that we would get $F=\nabla(Iab\cdot B_z)$. Now what surprises me, is that we're considering $B_z$, instead of the other two components of $B$, which are the ones that contributing to the downward facing force... And it seems to me that $\nabla(IabB_z)=0$, but of course that's not correct. Could anyone show me my error?

Quantumwhisp

21:24

heyho you lovely people: Does somebody know the Book "Quantum Field Theory I" by Edouard Manoukian? I have some questions about the 4th chapter, and thought maybe somebody could answer them to me

Semiclassical

"Your MATLAB license will expire in 02 days."

well that's annoying

enumaris

and back from lunch

danielunderwood

I'm still magically included on my university's MATLAB license

Semiclassical

i really should figure out how to use python

enumaris

it's ez

danielunderwood

21:31

I like it a lot better than matlab, though you have to import packages for most of the matlab functionality

Sadly I still feel like sympy doesn't quite compare to mathematica

alarge

You could always use Julia if you want an easier transition

(or Octave)

danielunderwood

Julia always sounds pretty neat, but I've never seen anyone using it

alarge

Yeah I think it came out too late. It would've been a solid contender had it come out a few years earlier as numpy/scipy were not as established

enumaris

octave is just open source matlab isn't it

alarge

Kind of, yes

enumaris

21:35

I used it for andrew's ML course

wasn't very impressed with it

I prefer to just stick with python

alarge

I was a heavy user of Matlab ~10 years ago when I then switched over to Python

Matlab's syntax is really unbeatable, and it's super fast as well

The functions are well documented, and from what I remember, well planned

Octave was much slower and from what I remember, it wasn't easy/possible to compile like with mex files in Matlab

I think Python has something called weave that I used to use a bit

But Python for mathematical scripting is definitely less ergonomic

And I would use neither language for any big project if I had the choice

Semiclassical

the main thing is that I need something open-source

enumaris

try assembly

I hear it's customize-ability is unmatched.

alarge

So you can use Octave, that's open source

as in Julia

Octave is not really drummed up anymore. At least a few years back when MOOCs were all the rage, a lot of the courses did use Octave (because presumably the on campus versions were Matlab based)

Julia's tried to get a foothold and I've seen them marketing to finance applications in particular, though I don't know if it's used in large scale production anywhere

enumaris

make sure the language doesn't have "dynamic memory allocation" if you want "production-level" "quality"

Semiclassical

21:45

thankfully, all I want to do is "compute some stuff"

(n-dim convex hull stuff, which matlab has by default but mathematica irritatingly doesn't)

enumaris

this might help: en.wikipedia.org/wiki/Comparison_of_assemblers

danielunderwood

22:07

python probably has most of the things you want to compute...you just have to find them

For example, convex hulls from points: docs.scipy.org/doc/scipy-0.19.0/reference/generated/…

and ideally you also have to hope that whoever wrote the code is maintaining it

Semiclassical

been seeing a lot of Qhull as i've looked around for implementations of convex hull stuff

What I'm specifically interested in, for sake of specificity

The intersection of a convex hull and a generic hyperplane should be another convex hull. So the problem is: given the vertices of a convex hull and the equation of the hyperplane, compute the vertices of the new convex hull.

which really seems like it should already have a solution

(and not just a solution in 3D or 2D)

Lozansky

22:25

@ShaVuklia Well, that's not an infinitesimal loop is it?

enumaris

sounds like there would just be an algebraic expression to get that solution

Semiclassical

maybe? i don't have a lot of sense for how well you can do this stuff

the algorithms that come to mind for me are quite primitive

enumaris

somebody smart probably figured it out some time ago

that's my default go-to

Semiclassical

yep

one bit that makes it tricky is that a common problem is evidently to find the intersection of two convex polygons

but that's not what I want. so there's a good number of false positives on google so far

enumaris

22:46

build an AI to do google searches better

danielunderwood

23:42

Seems pretty easy return filter(query.match, items)

I guess you better slap a sort around that too to get the best answers first

@enumaris do you know much about the ML courses amazon released?

I think I mentioned something a while back about mechanical turk to label datasets. Looks like they have something specifically for labeling now aws.amazon.com/sagemaker/groundtruth

enumaris

23:59

@danielunderwood I haven't looked into that

Transcript for