Hi.

I need to convince a referee that an experiment/theory we did is of "broad interest".

If anyone wants to give advice we'll add you to the acknowledgements section.

Link (pdf): web.physics.ucsb.edu/~martinisgroup/papers/Sank2016.pdf

If interested, focus on the introduction, abstract, and conclusions.

@ACuriousMind Still here?

@DanielSank Wait. You got a complaint from a referee about that paper?

@dmckee You're surprised?

It's an improved description of a problem that stands in the way of engineering large scale quantum computing devices. How much broader do they want the interest to be?

I have never not gotten rejected in a paper submission, ever.

@dmckee It's not just an improvement, nobody had any friggin' idea what was going on before.

I like the way you phrased that.

"problem that stands in the way of building a large scale quantum computer"... heck yes.

I suppose you can add explicit language to that effect.

Welcome to the acknowledgements section.

You wouldn't believe the feedback we get on our papers.

They're usually like this: "This paper is well written and relevant, but it's too specific to quantum computing (or whatever), so don't publish it in this journal".

Or "...but isn't of broad interest".

@DanielSank ::chuckles::

Every, single, time.

Uhg.

My paper on world-record speed and accuracy in solid state quantum state measurement was rejected because it was "engineering".

I general get "This paper is well written and relevant, but there is this long list of editorial changes and new figures I want to see".

(Which is related to why I lose my temper when people here vote to close applied physics questions as engineering... although that's gotten way better)

Which makes for another couple of weeks of scut work, but at least it presages eventual acceptance.

@dmckee Well heck, that's great though.

@dmckee yeah

I wonder, is the problem partly that the field is so new there is not a widely accepted notion of what-papers-go-where and how-do-we-talk-about-our-stuff.

I don't think so, but what do I know.

I know that in my field those are the kinds of things you can only learn from someone who has been in the fray.

But if different referees have different ideas about that you're in for a world of hurt.

If someone submits a paper where they use superconducting qubits to model a molecule or something, it goes right in (usually). However, all the important results that go toward making that possible get upturned noses.

I shouldn't say it goes right in... that's not true.

I mean, the abstracts you have to write now to get into big journals are just plain ridiculous.

web.physics.ucsb.edu/~martinisgroup/papers/Roushan2016.pdf

^ exhibit A

I dunno, that first sentence is just so Hollywood to me.

Maybe I'm complaining about nothing.

@dmckee Oooh that sounds terrible.

Switching topics...

@dmckee this answer of yours is.... wonderful.

@DanielSank That first sentence reminds me of something my junior would turn in before they had learned to write like a physicist. I hope we're not heading for that world, I'd have to take up drinking.

@dmckee Well, head to the liquor store, buddy.

I am right now making my paper worse, in precisely the way you're talking about, to get it into PRL.

I had a liquor store I really liked in Kansas. They got in odd-ball wines in the $10--15/bottle range and knew enough about wine to be helpful without being snotty about it.

Now only if more physicists knew programming we could use that metaphor more often.

Haven't found a place like that 'round here yet.

@dmckee Where's here?

Joplin, MO.

Not a bad town for it's size, but still pretty isolated.

Being underaged is a disability

@0celo7 But rarely a permanent disability. So no government money for you.

WHY DOES LINEAR ALGEBRA SUCK SO MUCH

Saying that in a roomful of physicists might not be the best plan. Especially as we apparently need to be drunken physicists.

@0celo7 I see you're using all caps today. Back on the ol' ignore list you go.

Just make up your mind already.

Hm, maybe that wasn't nice. I apologize for the wording, @0celo7.

Please do not show up and shout in the chat room. It's rude.

It's particularly annoying when the though represented by the all caps text is almost entirely devoid of meaning and offers no path by which anyone can enter into a meaningful discussion.

I am extremely confused by the order and signs in Poincare duality.

And the linear algebra notation is not helping

And of course various texts disagree

@JohnRennie Up late with cohomology again

You're up early with servers, again

What else is new

I've discovered one of my backups isn't backing up what I think it's backing up, which is slightly alarming ...

@JohnRennie Woah, back up.

You have a backup backing up something other than what you want to backup? So if if the server goes down, then once you get it back up and load the backup, you won't have backed up what you intended to backup?

@DanielSank This server has a large directory that is backed up using incremental backups based on the timestamp. So if a file on the server has a more recent timestamp than a file on the backup that files gets copied over.

This means only changes are replicated so we don't have to back up the whole, very large, directory every night.

yup

But, it appears that some files on the server have been changed, but have a 2015 timestamp. So they aren't getting backed up.

Yikes.

How this happened i don't know. I can only guess that they were modified by a client PC with the date set wrongly.

I'm suspicious of that explanation.

But that shouldn't be possible because Active Directory will complain loudly if a date is wrong by more than 5 minutes.

So I can't understand what has happened.

You tested that?

It's only a couple of dozen files out of a million or so, but even so it's worrying that it happened at all.

@JohnRennie Agreed.

Did you test that Active Directory complains about mis-dated files?

@DanielSank Yes, if the timesatmp on a PC is out Active Directory won't authenicate it and that user doesn't have access to shared folders etc.

Ok

How do you know the 2015 dates files were modified?

The 2015 files are different to the ones on the backup (that are dated 2016). But exactly what happened I can't tell.

@JohnRennie The backup has similar files (i.e. same name, similar content) with 2016 dates, but the files on the server have 2015 date?

Yes. On the backup the files have a Feb 2016 date, but on the server they are 13th Aug 2015.

Oh this is getting interesting. Does your system log writes to those directories?

@DanielSank No, the activity is too high to be routinely logged.

That 13th Aug 2015 date is suspicious - it's awfully close to being exactly a year ago.

@JohnRennie Indeed.

That's why I wonder if we have a host somewhere with the year set wrongly.

@JohnRennie I'm not familiar with Active Directory.

Fun puzzle, even if distressing.

@DanielSank it's just a directory authentication service like X400. It holds a database of known PCs and users and it authenticates connection requests from both PCs and users.

@JohnRennie Are clients writing directly to the server or do clients have a local copy which syncs with the server?

Clients write directly to the server. There is only one copy of any file, and that copy lives in a shared directory on the server.

Actually I have a suspicion re what might have happened.

The accounts software has recently been upgraded to a new version, and when an old version database is opened it is automaticaly updated to the new version.

I wonder if the automatic update process is writing incorrect timestamps.

It looks to me as if the affected directories are ones that have been used recently.

However it isn't all files in directory that are affected, just one or two of them from any particular directory.

Hey @dmckee we're playing "what's wrong with the file backup system?"

I'm going to do a complete copy of all the files overnight then a file compare to ensure we're starting with all files the same. That way if the problem recurs at least I have a known starting point.

Or probably "what's wrong with the shared file hosting system?"

@DanielSank less fun than physics :-(

Though it pays better :-)

@JohnRennie Depends on your fetishes and what physics you have in mind.

I'm terrible at windows stuff.

Windows shmindows.

The thing with Windows is that underneath the layer of fluff it's a really well designed and very robust OS.

I kinda get what Microsoft was trying to do with those protocols, but they always seem unnecessarily baroque to me.

@JohnRennie Then how come everyone hates it so much?

Non-(computer)nerds just see the appalling crap that MS layer onto the user interface.

@DanielSank Because the layers of fluff obscure the beautiful kernel.

And the embrace, extend, extinguish strategy has left them with a legacy of unnecessarily complex thingies to deal with.

@JohnRennie I'm a nerd and I dislike Windows for technical reasons. For example, I was writing a chat client/server once and I wanted an asynchronous loop. Guess what! you can't select on standard IO in Windows!

I'm afraid you can ...

I know nothing about the OS internals though.

If you're using sockets then select works just as BSD intended.

@JohnRennie Right but normal keyboard IO doesn't use sockets.

In Unix you can select on anything with a fileno, which is basically everything.

So you want to see if there's a keypress in the keyboard buffer?

@JohnRennie Yes.

Of course in windows you use some msvcrt lib call, like kbhit.

But that's just annoying.

That can be done because I've done it, though I would have to go root out some old code to remind myself how it's done.

I would be interested in how to accomplish that.

Once you're done figuring out why all your dates are messed up.

A keypress generates a message that goes into the queue of the window that currently has the focus. So you just need to look at the message queue.

If you want to interrogate the keyboard directly use GetAsyncKeyState or GetKeyboardState.

Right, the point I was making before is that this is all more... particular than in Unix where I just say select(file1, file2, etc.) and be done with it.

Ok, but I think you have to be careful to distinguish between the merits of different designs as opposed to the it isn't what I'm used to judgement that we're all prey to.

Oh, absolutely.

I am relatively inexperienced in all of this business.

The Windows kernel doesn't subscribe to the unix approach that everything is a file

On a related note, I hear that .NET is awesome.

@JohnRennie Right, I think this really throws people.

@DanielSank Partly that's historical, but partly there are good reasons not to follow that route.

But it certainly adds complexity.

Did you see my link to the firebase docs yestereve?

Looks like new life is breathed into it.

I remember when Firebird was forked from InterBase - many, many years ago :-)

Firebird is an excellent RDBMS, it's just that in the Windows world using SQL Express is a no brainer.

(aside: so-called "dynamic" languages make me said)

@JohnRennie I see.

John, are you experienced with SQL databases in any capacity? And if so, what is your tolerance for me asking you stuff about it?

I have taught myself the basics and am trying to now use it for Great Good.

You're welcome to ask, and I'll help if I can.

@JohnRennie I appreciate that.

I do have a question for you.

Fire away ...

I am setting up a database to keep track of microchip wafers we produce, and measurements of the resulting chips.

Each wafer is chopped into four pieces. Each piece cut into a number of so-called "die", and each die contains, say, one hundred little electrical devices.

I have already encoded this in a schema.

No problem there.

I now wish to write a client script which can take the results of a measurement run (a csv file) and upload the data to the database.

So here's the tricky part: each row of the csv file represents one measurement performed on one device, on a particular die.

To enter this data, I need to know whether or not the table representing dies has that die already entered.

If the die doesn't exist, I have to create it, and to create a die, I have to make sure the wafer piece it came from exists, etc. etc.

Is there a sane way to handle this problem?

Typically you code referential integrity into your schema.

Yeah I think I did that.

If my script tries to create an entry for a measurement on device X which comes from die Y, but die Y doesn't exist yet, it's an error.

So each row in the Pieces table contains a *wafer_id" field, and you cannot add a row to the pieces table unless a matchin g wafer_id exists in the Wafers table.

correct

Yes, I used foreign keys as appropriate.

I'm mostly asking how the client should figure out what to do.

So just do a select count(**) from Dies where die_id = xxx*.

If the result is zero then that Die needs to be entered.

I want to avoid doing that for every row though, right?

That's a lot of communication.

I feel like I need to work my way through the schema, checking that each "dependency" exists, and then after that upload the measurements.

But I'm not entirely sure what I mean by that.

I would simply attempt the insert and trap any errors.

@JohnRennie Interesting.

in many cases the insert will just work and no extra effort is needed.

If there's an error, back up a step?

If the insert fails than look at the error, and if it's a foreign key violation you need to add the Die.

Like, if the measurement entry fails, back up and try to enter the die. If that fails, back up and enter the piece...

@JohnRennie Right.

Ok.

Ok thanks that's a helpful hint.

Yes, I see how to do this.

I always wrap this stuff in objects. So I would have a CDie with a method to add itself to the database.

Oh that makes me feel good! I'm using a rather well designed ORM called sqlalchemy.

Then the code in CDie.AddToDB() handles all the checking whether the Piece exists.

I was just thinking I'd add methods for what you suggested.

And CPiece.AddToDB() handles the checking for whether the Wafer exists.

@JohnRennie Right.

I'm thinking of adding a dependencies list to each model class so that I can recursively check whether dependencies exist when there's a write failure.

@DanielSank I don't see the point of doing that as it's hard coded into your schema

A CMeasurement shouldn't know or care about Wafers and Pieces. All it cares about is its parent CDie. It should assume that the CDie knows what to do with any higher level objects.

Likewise a CDie only cares about its parent CPiece

@JohnRennie Well I was thinking of hard-coding the dependencies of each model...

By "model" I mean "class representing a db table".

It would mean that the methods I write don't have to be model-specific.

The model-specific data is just a list of dependencies.

Maybe we're talking about the same thing in slightly different terms, but a CMeasurement should care about a little as possible. In this case all it cares about is that its parent CDie exists, and nothing else about the schema matters to it.

I have to think about the code. Thanks for the suggestion!

@JohnRennie Agreed.

I think we have the same idea (well, the idea you suggested) and are imagining the details differently.

When your database exceeds a few million records then you have to start worrying about normalisation, and that's where life starts getting complicated :-)

@JohnRennie We're going to hit those numbers within weeks.

Should I be afraid?

We do 10,000 measurements per run.

@DanielSank Really, wow!

@JohnRennie Yeah, microfabrication for the win.

Do you know what I mean by normalisation?

@JohnRennie Yes.

It's sort of the database equivalent of the DRY principle... sort of.

@DanielSank No, not really.

I read a very long tutorial on SQL about a year ago.

It went through normal form in detail.

Suppose your Measurements tabe has 20 columns, and suppose that most of the time you're only querying for 10 of those columns.

@JohnRennie ok

As the Measurements table gets big it starts getting slow for the RDBMS to pick out just the 10 columns you want most often.

The slowness is just the speed the disk head can jump around reading the data off the platter.

So you split your table into two tables - one with the 10 most frequently used columns and a second table with the lesser used columns.

The documentary Particle Fever caused me to wonder about what happans if multiverse is interacting, random googling give me this sciencedirect.com/science/article/pii/S0370269316302465

By doing this you make your 10 column query faster because the RDBMS doesn't have to pick out those 10 columns from all the other columns.

@JohnRennie ok but I feel like we just introduced a problem.

The downside is you now have to manage the links between two tables for every measurement.

@JohnRennie Indeed.

Shouldn't I just put a foreign key in the first table which refers to a row in the second?

But if you're Amazon, or Google, or ebay, you have to do this sort of thing to make your vast databases run at a decent speed.

You get the speed but at the expense of extra complexity.

ok

@DanielSank I would just use single tables for now, and only worry about normalising if performance starts to be an issue.

@JohnRennie Right.

Wait, normalizing means splitting tables to make access fast?

I thought it was this

studytonight.com/dbms/database-normalization.php

To get the normalisation right you have to know how your data is being accessed i.e. what splits will help and what won't.

@JohnRennie Right.

To be honest I would get help from an experienced DBA at that point. Google must have some :-)

Indeed.

[Random thought] what happens if we taught an AI particle physics, and it will soon came up with some model that is so alien to anything us humans can think of

Then that would be awesome.

@Secret that's what neural nets do ...

(Cont.) What is alien, how to speak alien? We humans have a tendency to look for patterns. We love symmetry. But being one who denies symmetric approaches does not make us any more not pattern finding (we just try to find a pattern that is asymmetric, a negation of the concept symmetry). Will AI, which one day will be able to reach a self conscious state and hence became an independent living entity, be able to have a mindset that bypasses the limitation of humanity? (cont next)

You have computers came up with crazy mathematical proofs that will take some time to analyse by people

Can we ever speak abstract. The multiverse is utter randomness while SUSY is ordered, symmetric. There are various models in between. But are there something beyond that, something that because of our fundemental wiring, cannot be comprehended easily?

(Sorry I got a bit too touched by the documentary, thus the above are not really questions, but raw thoughts sprewed from the mind)

[Reading Richard Dawkins the magic of reality, thoughts] when seeing something seems breaking laws of nature, rather than saying that something that it is a miracle or magic and let it be, perhaps asking why and how it is a miracle or magic. Then if the result can be genuinely replicated in a systematic way, it becomes science

[Reading the time chapters of Brian Greene the fabric of the cosmos, thoughts] (to be discussed with sleeeah later) What if, the change in spacetime topology is not provided a mechanism in GR, but is in fact impossible because it potentially violates the conservation of information...?

sciencedaily.com/releases/2016/08/160817142602.htm How will this be related to charge density waves that propagates in superconductors?

@DanielSank Actually, I think I was talking bollocks about the performance issues and normalisation - I was getting mixed up with something else. Disregard anything I said about splitting a table making the queries faster.

[even more random thoughts] If magic is something that follows some inexplicable rules and some tangential cause and effect of seemly unrelated things; while science is a systematic way to achieve understanding by modelling and then test against reproducible unbiased results and methods (experiments). Then

Suppose we have found a phenomenon that is reproducible (omitting statistical noise like occasional fails or unexpected outcomes) but we found no underlying mathematical rules that govern the outcome, only a discrete set of cause and effect between a pair of concepts that has no common underlying mechanism, will we call this magic, or science?

Is "reproducible non mathematical governed nonrandom phenomenon" possible in the framework of physics?

(Cont.) Put it simply is it possible to have reproducible phenomenon without some underlying explanation...?

I i have a rotating disk in the x-y plane and i apply a force along the x-y plane in line with the center of mass of the disk , will the disk move along the direction of the force or along a different direction ? I am asking this in reference to the precession of a gyroscope. Thank you.

@ACuriousMind is that Freya from KotOR II?

@Sanya Kreia, but yes

Is someone in the mood to discuss gyroscopes?

well, it has been a bit that I played it, so I got the name confused :D but even though it is an amazing game, I have to admit your orc was nicer to look at @ACuriousMind

@Sanya Well, not everyone can be as handsome as Thrall ;P

:D even though WC3 never got me

Oh, so many hours that I sank into that one, it's not even funny.

Well, we all have those games :D for me it was e.g. Command & Conquer Generals - or both KotOR which I both played multiple times ... :D

KotoR was great, if only they had let Obsidian actually finish the second part instead of rushing it. :P

you know the mod that adds some of that content?

I've forgotten the name, but it was worth playing again with it

Yeah, the restored content mods are a must

Still, the end game is notably lacking compared to the earlier regions

Same with other great RPGs, like Vampire: Bloodlines; Amazing first half, then the studio went bankrupt and the second half is a horrible combat slog, mostly

yeah and I think the storyline and the characters in the first one are just a bit better - I mean, Taris in the first game has such an amazing atmosphere, nothing in the second one beats that

well ... I like the combat in kotor, so I don't mind excessively much :D :D :D

@Sanya Holy wars have been fought over which part is better ;D

@ACuriousMind which I can even understand - because gameplay-wise, the second part has added some nice features; it's just the story, atmosphere and the (in part 2 lacking) completeness which are really better in part 1 ... A bit like GTA Vice City and San Andreas :D But well, I will just remain boring and diplomatic saying I like both of them :P

I actually like the characters in the second part better, though I'm inclined to agree with story and atmosphere

Bastila? Where is a character as cute as Bastila in part 2? :o Oh, and Mission and Zaalbar are pretty cool as well

Oh god, Bastila is such an annoying self-righteous brat :P

Though well-written enough that I actually dislike her :D

I find her pretty adorable :D

that self-righteousness is what makes her cute :D

I do find the conversations with Kreia and Mandalore very interesting, and HK-47 is just always hilarious

hk is cool, yeah

I'm pretty bored with Mandalore, I have to say; Kreia doesn't interest me that much either :| I mean, all that pseudo-deep talk about never being right ...

@Sanya It is "deep" for the Star Wars universe, where usually everyone just clings to one of the goody two-shoes or evil for evil's sake viewpoints.

And you have to keep in mind that everything she says might just be to manipulate you, not what she actually believes

@ACuriousMind well, yeah - but that's what Star Wars is for, after all - nice black and white categories

@ACuriousMind grant you that - but it doesn't work well on me because most of the time, I'm just not paying enough attention :D

lol

I guess we're playing games for slightly different reasons ;)

whom I do like in the 2nd part is Handmaiden, that's almost a bit of character development

well, depends ... I do read the dialogues, but I never got interested enough in Kreia because she was quickly annoying me

@shashank I have to go out, but I'll be back in a few hours if you want to discuss gyroscopes.

Ok, @ACuriousMind . Did you hold the thought?

@Sanya Agreed

@0celo7 Kind of? Something something Poincarê duality

@ACuriousMind Ok...so is the Poincare duality condition $\int_S\omega=\int_M\eta\wedge\omega$ or $\int_S\omega=\int_M\omega\wedge\eta$?

here $S\hookrightarrow M$ is a closed submanifold.

and $\omega\in H^k_c(M)$.

@0celo7 Either choice of sign should be fine.

@ACuriousMind But not when both are used in the book...

@0celo7 Are you trying to get me to debug notation I haven't chosen or read again?

No

The isomorphism of Poincarê duality is not supposed to be canonical in some sense, so the choice of sign doesn't matter, you just need to be consistent during a single application

@ACuriousMind Ok, ok.

My real question is...I have a compact manifold $M$ and a linear map $f^*$ on the cohomology.

I want to take the trace of this map using Poincare duality

To get a trace I normally need something like $\sum \omega^i(fv_i)$ where $v_i$ is a basis and $\omega^i$ a dual basis

I'm not sure what you mean by "using Poincarê duality"

So...how do I do that with Poincare duality?

@ACuriousMind Argh, I have to get ready for class...please hold the thought again

I hate school

The trace of a linear map is defined simply by taking the trace, I'm not sure what Poincarê duality is supposed to have to do with that

Because Poincare duality gives a dual basis to the forms in cohmomlogy!

or however you spell that word.

@ACuriousMind Is it not true you can use a dual basis to take a trace?

Sure

If $\{v_i\}$ is a basis of $V$ and $f:V\to V$ is linear then $\omega^i(fv_i)$ is the trace if $\{\omega^j\}$ is a dual basis.

I even managed to prove that

So I have a basis of $H^\bullet(M)$ and can use Poincare duality to find a dual basis

then I construct the matrix elements and sum over the diagonal

See, what I don't get is why you want to make your life complicated by using the Poincarê duality to write down the dual basis "explicitly". Just express $f$ as a matrix and take its trace

@ACuriousMind ...

Just trust me.

I will work on my problem some more

But sure, since Poincare duality tells you that $H^{n-k}$ is dual to $H^k_c$, you get that there's a dual basis

What is the problem?

And when it is done you will see why I am doing it this way.

@ACuriousMind Minus signs.

The minus signs are crucial to my proof.

...

What am I supposed to do about that?

@ACuriousMind Nothing right now.

this:

cheers

made my day

@Jim Your day appears easy to make :)

@ACuriousMind I live a happy life

@ACuriousMind Yes, the signs work out.

I have completed the proof.

Now, to type it.

@ACuriousMind "Poincaré" Shirley

@0celo7 I still don't understand what part of this I could've helped it or why it was important I hold a thought

@JohnRennie Haha, no idea why I pressed the ^ there consistently

@ACuriousMind I have to admit I had to Google it to check :-)

@Jim: I do, I do (need the rep). I'm within spitting distance of 200k but the last few days things have slowed to a crawl! :-)

@JohnRennie your crawl would be my great day

with over 3000 answers, how could you not get at least 50 to 100 rep a day?

my mistake, over 4000 answers

I'm sure I would eventually hit 200k even I completely stopped answering. However the site seems quiet at the moment (maybe everyone is on holiday) and legacy upvotes have slowed to almost nothing.

Why does this question have 3 upvotes and 4 favourites from just 11 views? Is there something impressive about it that eludes me?

Awesome question, truly a work of art

A work o' fart indeed.

lol

I've seen Americans write fibre bundle.

Why?

@ACuriousMind chill.

I'll write the proof when I get to my office and then you'll see.

@0celo7 Because they are Americans who have learned to spell? :-)

@ACuriousMind I needed your help for this computation

and I think it's correct.

IT WORKS

@DanielSank I hate linear algebra less now :)

@0celo7 You...needed me to tell you "pick a sign convention and stick with it"?

@ACuriousMind Yes.

woooooooo