Wolfram Mathematica

@user21 Are you in my local time zone? Holy sh**, didn't expect you so late :)

Murta

Someone want to have fun and help Brazilian democrary? :)
http://community.wolfram.com/groups/-/m/t/824675

@Murta Absolutely impossible to turn this into something accurate. In the far end of the street, you wont even be able to have 1 pixel per person. Image processing is the wrong way. Why don't you approximate the street width and length and estimate how many people per squaremeter are in such a dense crowd? This will give you better results, I'm sure.

1:52 AM

@Murta, if you'll take up halirutan's suggestion, I'm sure you can get information on the street length/width from your infrastructure department.

@J.M. Or google maps.

anyone knowledgeable on RBF Networks by chance?

RBFs, yes. RBF networks, no.

2:08 AM

in RBF equations, is the value 'e' the mathematical constant 2.71?

@Woodrow Depends. Where did you see this formula?

hm this is weird. In a book I have that glosses over RBFs, the equation for Gaussian RBF is phi(r) = e^(-r)^(2), however in wikipedia: en.wikipedia.org/wiki/Radial_basis_function the gaussian equation is phi(r) = e^(-r * some_weird_curvy e)^(2)

@Woodrow That $\epsilon$ just tweaks the width of your Gaussian.

I see, that makes sense, since my book talked about making use of RBF width, but didn't explain where it fit in. More errata I suppose

Thanks. and I presume the normal e is 2.71?

@Woodrow Yes, the usual exponential constant.

2:22 AM

thanks

Vitaliy Kaurov

@Szabolcs should be fixed, can you check? Thanks for the message.

@Kuba should be fixed, can you check? thanks for the message

yode

2:38 AM

{2.6135182051634342, -2.6690957180727297, 0.26060239174734157, -0.4182243295792278}

I want to gather this number

every group have two element and its diffrence is near 3.14

how to do this?

every group have two element and its diffrence'absolute value is near 3.14

user21

5:12 AM

@halirutan, naa, im a a few hours ahead of you ;-) so it was early morning for me already

Hmmh... is this use of Band actually considered supported functionality?

Range[5] + SparseArray[Band[2] -> {a, b, c}, 5]

(* {1, 2 + a, 3 + b, 4 + c, 5} *)

It's not immediately obvious from the documentation of Band.

6:48 AM

@MichaelHale it looks like a reef shark. where were you?

@Kuba sorry for the late reply. I am not on here much. I think it is an example of Mathematica having several ways of doing things. It can be useful for making code a little bit briefer

9:08 AM

@VitaliyKaurov i.stack.imgur.com/URyzV.png

Q: Handling Kernel start. What is a full initialization stack and what place the Autoload has there?

9:32 AM

0

According to: tutorial/WolframSystemSessions Initialization On startup, the Wolfram Language kernel does the following: Performs license management operations. Runs Wolfram Language commands specified in any -run options passed to the kernel executable. Runs the Wolfram Lang...

Szabolcs

9:46 AM

@yode Not heavy :-) Nothing to do with your post

10:27 AM

@MikeHoneychurch No worries. p.s. I don't agree :) the point is with this approach you have to put Appearance in each Button or create custom function to invoke it.

11:09 AM

I have been playing with the "Wolfram Development Platform", that is the web notebook interface, free tier. It would seem it doesn't really run out of computing time any more, at least easily. (It might be it throttles CPU cycles, though.)

Which makes even the free variant surprisingly usable.

Now that I said it, it "exceeded the plan" on one computation. But really, that took minutes.

11:28 AM

@kirma What were you trying to do that it cut off?

11:43 AM

@J.M. Nothing particularly memory-hungry, but just computing thousands of integrals with varying parameters.

11:55 AM

@halirutan Hi, do you have a minute?

@Kuba I guess I can spare a minute, yes. What up?

@halirutan What people do when the work from different machines and they want to have access to files they don't want to track with git?

I'm using google drive. Simple as that. For Windows and Mac there exists a nice apps. On my Mac, I just get a normal local folder which is then automatically synchronized with the cloud and kept up to date.

With drive, I can share files and folders with colleagues and give them read/write access. So for instance, I have a publication folder for one of my projects containing all manuscripts that are important for this. Colleagues can just drop other publications there or get them. This is much better than sending everything with email or skype.

@halirutan I see, thanks for the answer :)

@Kuba np

12:21 PM

@MikeHoneychurch Yes, I was in the Maldives. That was the baby reef shark that swam around the resort. There was also a baby nurse shark that would occasionally come by. I saw 4 larger reef sharks while I was getting my scuba certification. All were smaller than 2 meters though. I didn't see any as big as in the Belgian guy's photos. He was on a dive focused trip though. Stayed a full week on a traveling yacht and went diving 3 times a day.

He has a video from a night dive they did where dozens of larger nurse sharks swam around them. It looks straight out of National Geographic. I just used a disposable underwater camera. The colors were extremely washed out in many of my photos. Had to play with the levels a lot in GIMP just to make the shark recognizable in mine.

Gave me an appreciation for how much auto focus and exposure adjustment we take for granted even with a camera phone these days.

I was mainly there to try one of those "villa on the water" places. I recommend it.

I think the only thing I got a picture of that the Belgian experienced diver didn't get a better picture of was a moray eel.

More pics here (above and below water): dropbox.com/sh/vdswrgxl4sx40ne/AABiv-lh4Hfe2C_TcOgil3cwa?dl=0

Q: How to prevent $\text{Quantity}$ calculations from converting units

12:45 PM

@Kuba Kuba - if the machines are all either linux or mac you can mount the remote directories easily using sshfs

Edmund

1:59 PM

Since posting a question I have an almost working answer to it. Should I post this as an answer to the question or edit the question and add the almost answer to ask for help completing it?

6

I have an accrual calculation that I would like to preserve the units even though they are of the same dimension (e.g. time). For example, I would like the following Quantity[5, "Days"] / Quantity[52, "Weeks"] to return in units of days per week instead of a unitless number. Later in the proce...

units upvalues

Shutao TANG

2:44 PM

user3183724

If I have a list of points A = {1,4,7,2,4,1,2,9} and I want to (list)plot them all at the same x coordinate B = 3, is there an easy way to do that? My issue is actually that for each x coordinate A has a different length, so I can't just make a table. So more generally I want to plot A1 = {1,3,4,2,1} at x = 2, A2 = {1,5,2} at x = 3, and so on.

Try something like this,
`A = {1, 4, 7, 2, 4, 1, 2, 9};
A2 = {1, 5, 2};
A3 = {-7, 2};
Blist = {3, 4, 8};
ListPlot@MapIndexed[
Function[{a, index}, {Blist[[First@index]], #} & /@ a], {A, A2, A3}]`

3:02 PM

2

Accidental art...

user3183724

@JasonB Hm I think my example was confusing, so I'll try to be more clear. I have a function that outputs a list A of variable length, as a function of x. I want to listplot A(x) versus x, without knowing beforehand what the length of A(x) is at each point x

Edmund

@kirma Code?

Plot[Table[((2 Reverse@IntegerDigits[n, 2, 6] - 1) 2^-Range@6).Erf[(x - Range@6)], {n, 0, 127}], {x, -1, 7}, PlotRange -> All, Evaluated -> True, Axes -> False]

Slightly souped up for chat purposes...

Edmund

@kirma Nice!

A[x_] := RandomInteger[{-5, 5}, RandomInteger[{1, 5}]];
Table[{n, #} & /@ A[n], {n, 5}] // ListPlot

3:07 PM

I was trying to write code where all the points on the left side would start from the same location, but forgot to add a constant term to Erf... so it became a bit stranger.

@kirma That sure is an attractive graph!

Does anyone have experience with FindClusters?

I have these coordinates, and there is one clear outlier,
data = << "http://pastebin.com/raw/RGZ6iusm";
ListPlot@data

And I would expect FindClusters[data] to split it into 2 groups, but it can't detect that there is one outlier

3:31 PM

Even fancier I guess.

R. M.

3:50 PM

@halirutan This is mostly correct. The reason you wouldn't use pygments for a "new rendering per request" kind of setup on the server is so that you don't waste time and server cycles to do highlighting for every user. Javascript is simpler then because it is done client side, so less load for your server.

On the other hand, if your CMS is smart enough to do a "render once and then serve many until it changes" then you could hook up pygments to it. It would do the rendering the first time and then serve the cached page until your underlying .md files change.

The real question is, if you don't have any dynamic pages (e.g. time/location specific rendering) or things that need a database (users/interactions), then you probably don't need a heavy setup. Jekyll/Octopress/Pelican is probably the best solution. You still write everything in markdown and then the program compiles everything into HTML and CSS files locally. You simply upload these locally rendered files to a web server when you're satisfied and serve those.

The files that are rendered will only change when you make a change to the md, compile it and upload them to your server. This also makes it easy to put everything under version control.

@JasonB Your data has two copies of every point, but I can create the clusters you want with the following: FindClusters[data, 2, Method -> "Agglomerate"]

@MichaelHale Thanks - I had tried each of those separately (specifying number of clusters, or the method), but not together

4:15 PM

@R.M. The main reason why I thought it is a good idea to do this was, because with Pico, Grav and friends, you always load the md files to the server and they interpret them on the fly.

I was having a late night session with Pelican yesterday.

It is very nice too. The support for Pygments is shitty, because although it works out of the box, the inclusion of your own set of mma.css was difficult because there is no way to mark the code boxes so that the html-classes language-wl are used.

The only way to solve it was to bring the language-wl css class into the main site template which means that all codeblocks will have this class :(

Hey guys, I'm having trouble understanding something about using Accuracy or PrecisionGoal with NIntegrate, and was wondering if anyone could shed some light on it

I'm integrating a function over two dimensions. The function has a peak that should be easy to integrate for a small area of the integration region, but as soon as you go away from that peak, the magnitude of the function becomes insanely small (10^-8400 size)

Because the peak is so much bigger than the areas around it, when I integrate, I don't really care much about the contribution of the areas around the peak, just the peak and a little area around it

It doesn't really seem to help setting the AccuracyGoal, it still gives me the same slwcon error

So how can I integrate over a region where a lot of the region is insanely small and should basically be ignored?

4:33 PM

@YungHummmma Did you try Method -> "AdaptiveMonteCarlo" or Method -> "AdaptiveQuasiMonteCarlo"?

I have never tried NIntegrate with Monte Carlo myself, but Monte Carlo seems appropriate for this task.

4:55 PM

@Pickett, I'll try that

@Pickett it...magically worked

what.

I need to go now for a bit, but here's my function in case anyone wants to see:

fdfn = 1/(1 + Exp[100*(x + y - 3)]) - 1/(1 + Exp[100*(x + y - 2)])

fairly simple

There is a Plot3D of it, you can see that its "peak" goes up to 1, but at x = y = 5, the value is ~10^-308

sorry, 305

the monte carlo thing seems to work though...

6:29 PM

I'm glad it worked, then theory and practice are in harmony.

7:11 PM

@Pickett thank you for the advice

I'm still trying to understand it though, which of the the slwcon things was happening with integrating my function?

(oscillating, 0 value, etc), because it doesn't seem like any of them are happening...

waaaait a minute... if I reevaluate the integral with Method -> "AdaptiveQuasiMonteCarlo" repeatedly, the value changes, fairly significantly...

oops, sorry, I just meant Method -> "AdaptiveMonteCarlo", no quasi

7:37 PM

@kirma what parameters did you change for your second plot?

Plot[Table[((2 Reverse@IntegerDigits[n, 2, 6] - 1) 2^-Range@6).Erf[
    2 (2^Range@6) (x - (1 - 2^-Range@6))], {n, 0, 127}], {x, 0, 1},
 PlotRange -> All, Evaluated -> True, Axes -> False]

Stuff inside Erf, basically.

7:59 PM

@MichaelHale difficult to get good colours in the water unless you are very shallow.

@MikeHoneychurch Yeah, I read you need a flash underwater if you want reds to come out right.

@MichaelHale flash can be a problem. reflections off particles that you otherwise do not see. You need very clear water for a flash

8:42 PM

@YungHummmma It's a probabilistic method so it's not going to give you the same result every time, but it does converge towards the solution so if you have enough samples you should still be able to get the precision that you need. Try increasing MaxPoints and see what happens, other than that I don't know what can be done since this is a very high level implementation of Monte Carlo integration, and the details can't be configured.

@Pickett, thanks for the advice

np, I hope it works out.

So I think I'm becoming more confused about this...