Wolfram Mathematica

hhh

00:22

Moved here.

Jacob Akkerboom

01:04

@Mr.Wizard ufff I am so confused. I would be very interested in your thoughts on my question. But maybe I should clean it up (for the bazilionth time) again first. Goodnight!

Oleksandr R.

04:19

@hhh please do not ask/pre-announce your questions here in the chat. This has been specifically mentioned by some people as being an unwelcome behavior.

Mathematica StackExchange (meta)

05:43

0

Q: @insert_name doesn't always work

In the comments sections of questions I often try to write "@whoever" before commenting. Sometimes it works fine; other times the entire "@_" statement is omitted from the comment when it is published. Why is this? Am I doing something wrong (some of the time)? Can it be fixed?

bug

Mr.Wizard

06:57

Is anyone here?

Hello @Michael

Yves Klett

@Mr.Wizard g´morning!

Mr.Wizard

Hello @Yves :-) Do you have time to run some code for me?

halirutan

@Mr.Wizard What's up?

Yves Klett

lemme see... yup :-)

Mr.Wizard

heh, not all at once! ;^)

Please, and thank you:

intervals[a_List] := {a[[Prepend[# + 1, 1]]], a[[Append[#, -1]]]}\[Transpose] &@
  SparseArray[Differences@a, Automatic, 1]["AdjacencyLists"]
f1[a_List] := (Pick[a, #, 1] & /@ {#~Prepend~1, #~Append~1})\[Transpose] & @
  Unitize[1 - Differences@a]
f2[a_List] :=
 Pick[{a, a}, {#~Prepend~1, #~Append~1}, 1]\[Transpose] & @ Unitize[1 - Differences@a]
a = Delete[#, List /@ RandomSample[#, 15000]] &@Range@1*^7;
timeAvg @ # @ a & /@ {intervals, f1, f2} // Column

Duh.. wait!

First need this:

Yves Klett

07:10

The Duh.. wait part causes an error, though.

Mr.Wizard

SetAttributes[timeAvg, HoldFirst]

timeAvg[func_] := Do[If[# > 0.3, Return[#/5^i]] & @@ Timing@Do[func, {5^i}], {i, 0, 15}]

hehehe

Yves Klett

0.084241
0.224641
0.215281

with 9.01 on Win 7 that is.

Mr.Wizard

Interesting. The SparseArray method is still more than twice as fast. Do you think f2 is more easily understood enough to include it in my answer along with intervals?

Yves Klett

the more the merrier

Mr.Wizard

@YvesKlett Is it more easily understood or are both quite opaque?

Yves Klett

07:18

hey, I am just running the code :-P... but yes they are both quite opaque to me

which does not tell you much since I still need my first cuppa

Mr.Wizard

hehe, okay, thanks

Yves Klett

going to get one now...

ugh, gotta rinse my cup first, not a nice sight

no wait, switching to my WRI alpha mug.

Michael E2

10:12

@Mr.Wizard Hi. I was asleep.

Rojo

14:13

Mornign

Jacob Akkerboom

Morning :)

J. M.

@Rojo Hola, buenas noches.

Rojo

@J.M. Buenos días J.M, tanto tiempo

J. M.

@Rojo Hay, mi casa es muy caliente...

2

Damn, it went to 38 today.

Rojo

Hahaha

No air conditioning?

J. M.

14:20

That's the tropics for you. Also, I'm too poor to afford air-conditioning.

Much less the electric bill after using said air conditioner.

Now, it's night, and I'm still sweating out the water I drank a while ago.

And how's your fine morning?

Rojo

@J.M. Aww

My morning's great. I didn't hear my alarm clock so I woke up better slept and got to work 1h30 late

Wonderful day, sunny, 20 degrees

J. M.

@Rojo I am allowed to be jealous, no? :)

(i.e. you should be thankful that "sunny" there is the "nice" kind, not the "furnace" kind.)

@rm, I wasn't able to get to that remote computer today. I might be able to send you or Szabolcs something tomorrow, tho.

Rojo

@J.M. You are allowed to. But this is actually the best time of year in my opinion, it doesn't last

rm -rf

@J.M. Ok, cool

rcollyer

@J.M. yes, you are allowed to be jealous. We had snow on Saturday.

Rojo

14:34

January in Buenos Aires, without air conditioning, hard to survive

rcollyer

@Rojo eww.

Rojo

@rcollyer Niiice

J. M.

@rcollyer Awesome. :)

rcollyer

@Rojo I don't live that far north, but the weather has been a little screwy lately. :P

Rojo

@J.M. and when you need the air conditioning the most, everyone does, so it's not unusual to get a powercut

J. M.

14:35

@Rojo Oh, that happens here too. :) Everyone has the same bright idea, and boom.

Rojo

@rcollyer It's the global freezing

or was it warming?

rcollyer

@Rojo :)

J. M.

@Rojo "Global Climate Screw-up" is what I personally use.

rcollyer

@J.M. when I lived in Baton Rouge, we were next to one of the more reliable substations in the city. We rarely lost power, but the neighbors across the street did, frequently, as they were on a different part of the grid.

Incidentally: My latitude - 43.15 N.

J. M.

@rcollyer Yeesh, nothing brings jealousy like "the house in front has power?! Fuuu..."

rcollyer

14:39

@J.M. very true. But, they were mostly shops, so not that much jealousy.

@J.M. Hope things cool down for you. Got to run.

J. M.

@rcollyer See ya.

Rojo

@rm-rf Hey, did any ideas on how to detect those roundy animals in the photo?

rm -rf

@Rojo Hey, I played with it a bit, but I wasn't able to satisfactorily connect pieces that should be together and not clump together those that shouldn't... a pretty tricky problem for a mere enthusiast like me :P

I do have some ideas though, but it's going to be a little work to translate to code

Rojo

@rm-rf Thankx. I'm open to hearing your ideas and do the work myself of course

Even a half manual approach in which you cilck on each being and they get automatically "rounded" would be somewhat helpful. I've got (so far) about 2000 images and rounding them by hand is monkey work

rm -rf

@Rojo My main idea was to use component measurements to "group" nearby fragments such that the major axis/minor axis was more or less close to 1. If you take a whole new blob, then it blow up because it's now an ellipse... the opposite if you leave big chunks

Rojo

14:56

a whole neeww bloob, a new faantastic pooint of vieww

Sigh, don't wanna work

Ok, SLAP, enough. I'll give that a shot, thanks

rm -rf

15:08

@Rojo I guess that is work related or otherwise, not an image you should be posting publicly?

If not, nikie might have a clever approach for it

Rojo

@rm-rf Work related. I'll check and if it's alright make it into a question to see what nikie or others come up with. I think there'll be no problem

rm -rf

@Rojo I think it's a pretty interesting question... IIRC, belisarius had answered a few like this on SO

belisarius

15:24

@Rojo A link, pls

@J.M. Ay! Hace mucho calor en mi casa!

Rojo

@belisarius ¡Ay! ¡Hace mucho calor en mi casa!

There are still 2 or 3 of people in the world that use ¡¿

belisarius

@Rojo Purists!

@Rojo Which "rounded animals"?

Rojo

@belisarius emailed

belisarius

ok lets see

Rojo

Can't upload at stack.imgur blabla, proxy "%$#

belisarius

15:30

@Rojo mmm are you using a corporate email thing?

@Rojo Seems your mail got lost in at some corner while trying to cross a bicisenda

J. M.

@belisarius Si, si, maestro... :)

belisarius

@J.M. doctore, please.

J. M.

@belisarius Sorry, title trouble. ;)

belisarius

@J.M. Oh, no! doctore is Latin. I love Latin. It sounds so uppish

J. M.

15:46

@belisarius Yes, I have to agree...

belisarius

@J.M. Exactly "Or it's just because Latin makes you feel intelligent and badass, with all the scientific terminologies and ... "

Oleksandr R.

@halirutan it turned out 3 times faster with POPCNT, in case you were wondering. Incidentally, do you (or anyone) know if there are any alignment guarantees for the array underlying an MTensor? The start of it should be 16-byte-aligned for optimum performance using vector instructions.

Rojo

16:02

@belisarius gmail

@belisarius and I sent it to the email in your profile

belisarius

@Rojo Something went bad, it hasn't arrived

@Rojo Have you used a $.10 stamp?

Rojo

@belisarius Check now

belisarius

@Rojo Here it is. You forgot to press "send" ... don't tell me

:)

rm -rf

@belisarius I wouldn't be surprised... he emailed me without an attachment :D

J. M.

@belisarius No, he mistook "save" for "send"... ;)

belisarius

16:04

Oh well, poor guy

J. M.

(is there no other short word that doesn't start with the same letter?)

belisarius

@J.M. store

rm -rf

@belisarius That's quite a different letter there...

belisarius

@Rojo I am having some difficulty in understanding what are you calling a being. Could you superimpose a few colored lines marking the desired result?

I mean ... joining the "beings"

Rojo

@belisarius Sure, a min

belisarius

16:14

@Rojo something like

J. M.

...are those algal cells?

belisarius

He broke down his girlfriend's necklace and now he's trying to put it together again

Rojo

@belisarius :S, can't load the image, proxy crap, doesn't like imgur

belisarius

@Rojo Is mine a good approximation?

rm -rf

@belisarius I think this is more like what he wants:

with the main difficulty being correctly identifying the different blobs shown by the orange circles and correctly grouping "possibly disjoint pieces" (in a binarized image) in the green circles

Rojo

16:26

This is frustrating. Image not found :P. I emailed you both @rm-rf and @belisarius

rm -rf

ha, so it's a combination of what we both showed :D

Rojo

@J.M. Skeletonema costatum

belisarius

@Rojo Yeah received ... but so ... 1) "beings" are unilinear, no branches 2) All the cells in a being have approx the same size

3) beings are almost "straight", no acute angles

Rojo

@belisarius 2) yes but getting the approximate size is part of the objective. In fact, I wouldn't mind if some of them are missed, just as long as they are not missed in a clearly biased way such as leaving out the biggests

J. M.

@Rojo Oh, a diatom! Neat!

belisarius

16:32

3) beings are almost "straight", no acute angles <--- ???

Rojo

@belisarius 1), 3), yeah, you can assume that

I think

belisarius

@Rojo Ok, what is the "final result"?

What do you want to measure?

Rojo

Most importantly, the equivalent radius

belisarius

@Rojo Of what?

Rojo

which is automatic with ComponentMeasurments once you have them all properly componentized

See the image I sent with the beings rounded in red?

Well,t he equivalent radius of those red regions

The radius of circles of same area

Dominic Michaelis

16:36

How can I force Mathematica to use the trig polynomials i want ?

belisarius

@Rojo One by one, the mean ?

Rojo

@belisarius Right

J. M.

@DominicMichaelis Exactly which "trig polynomials"?

belisarius

@Rojo Right what? One by one or The Mean?

Dominic Michaelis

I have the command (Integrate[#, x] &@ Sin[x]^2) // D[#, x] &

well i just integrate and differentiate sin^2(x)

Rojo

16:38

@belisarius Better to have the list of all, then the mean is automatic. But, if there's a simpler way to get just the mean I'm also interested

J. M.

@DominicMichaelis What does TrigExpand[] do to the output?

Rojo

@belisarius No

I actually need them all

now that I think about it. Sorry

belisarius

@Rojo I can put all the cells in a bag and call Baez to weight it

Dominic Michaelis

and i would like that mathematica gives me the result sin^2 (x) @J.M -2 Cos[x] Sin[x]

or wait

Rojo

@belisarius Haha

Dominic Michaelis

16:39

@J.M. yeah it gives -2Cos[x] Sin[x]

Dominic Michaelis

16:51

@J.M. TrigExpand[] doesn't change anyhting

J. M.

@DominicMichaelis Then, you are supposed to end up with something that can be transformed into $\sin^2 x$, yes?

Dominic Michaelis

oh I see my problem forgot some parenthesis

I did square the integral of sin not takingthe integral of sin^2 ...

now a simple simplify does the job

belisarius

17:10

@Rojo yOu here?

Rojo

@belisarius Yea

Dominic Michaelis

@J.M. thanks :)

belisarius

@Rojo Try this

i= Import ....

id = ImageDeconvolve[i, GaussianMatrix[10],
Method -> {"RichardsonLucy", "Preconditioned" -> True}]; ft =
FillingTransform@
DeleteSmallComponents[Binarize@GradientFilter[i, 20] // ImageAdjust,
30];
ImageMultiply[i, ft]

Rojo

@belisarius The ImageDeconvolve seems like a great idea

belisarius

@Rojo $10

@Rojo We could also fix it with 100 upvotes per month for a year

Rojo

17:25

@belisarius Haha, can I pay you with the 100 rep that MrWizard owes me?

belisarius

@Rojo I don't take dirty rep

@Rojo Anyway, Is that useful?

Dominic Michaelis

why does union don't sort exact values ?

Rojo

@belisarius It helps, definately. But the fight is far from over until I get the disjoint regions

belisarius

@Rojo Oh, of course. It's just a little step forward

Rojo

@belisarius Oh, if that step is little for you, please give me a hint

belisarius

17:28

@Rojo :D Is that homework?

Rojo

@belisarius It's homework I'm giving you :)

belisarius

@Rojo Oh, my teacher advised me to stick my nose at my own homework

Rojo

@belisarius Your teacher is old

belisarius

@Rojo ok. let me try

Dominic Michaelis

is it ok to ask another question ?

Rojo

17:31

@belisarius Just a hint of what you would try is helpful enough

@DominicMichaelis Ask. The worst thing that can happen, we ban from the internet

Dominic Michaelis

Why does Sort[{0, Pi, 1, 5}] gives {0,1,5, Pi}? I guess it is because Pi is not a number but well i think it is still unexpected

Rojo

@DominicMichaelis That's right and that's right

Mathematica has an order for everything, not just numbers

Dominic Michaelis

I used SortBy[{0,Pi,1,5}, N[#]&] to get what i wanted

Rojo

You can just write N

or Sort@N@{0, Pi, 1, 5}

no, but you lose your Pi there

SortBy[{0, Pi, 1, 5}, N] seems nice

Also, Sort[{0, Pi, 1, 5}, Less]

or #[[Ordering@N@#]] &

Dominic Michaelis

With Less he uses it's intern ordering which gives the same as onl Sort in version 9.0 at least

Rojo

17:49

@DominicMichaelis It shouldn't, Less does a numerical comparison

Sort[{0, Pi, 1, 5}, Less] should give {0, 1, Pi, 5}, and in my 9.0 does that

but your SortBy should be faster

Dominic Michaelis

Oh funny. Sortby[{0,Pi,1,5},Less] gives {0,1,5,Pi} and Sort[{0,Pi,1,5},Less] gives {0,1,Pi,5} I used the Sortby

Dominic Michaelis

18:02

for my problem my sortby is about 10 times faster as the sort with less

kirma

18:33

@DominicMichaelis You are not supposed to use SortBy like that...

Transcript for