Loglet in the Clouds

Dr. Modis created an iPhone app for s-curve analysis. Cool.

My javascript logletlab 3.0 app will probably work great on the iPad (it used to crash the iPhone). And its free…

Per a suggestion of Dr. Modis, The first logistic pulse is now a four parameter logistic:

where d is a initial displacement value.
The default constraints on d are zero and zero, so the
default is to fit a three parameter logistic for the first pulse
(i.e. the guessing algorithm will not attempt to guess a displacement)

I have written some preliminary and hopefully helpful documentation. Click “Documentation” in the tab bar above this post.

Comments appreciated.
Please send email to p e r r i n m e y e r _at_ y a h o o _dot_ c o m

I figured out the javascript to allow a user to copy data from two columns of a spreadsheet and paste it into the application. I have tested it on google chrome (fastest! - my recommendation ), safari (almost as fast, and firefox (about half the speed).

For this reason I consider the app slightly useful, and so I have decided to call it “beta.”. Click on the “Try LogletLab 3.0″ tab on the top of this page.

I thank the coders at http://stackoverflow.com for their helpful answers.

google has (finally) released an alpha version of their browser for mac and linux. And, as expected, it runs the alpha logletlab javascript very fast, more than double the speed of firefox (thanks to the v8 javascript engine). So I can now start on making a google spreadsheet logletlab 3 app. cool.

My first bug report!! thanks to jhelm! (user data not working). Fixed too.

I found a nice minimalist WordPress theme, fluid-blue, and I changed the defaults to make it even more minimal.  I think that I have latexrender set up so that it is at least readable.  For example, in our formulation of the Logistic growth model, we use to represent the growth rate parameter, but it has a different scale than usual.

And now I can write LaTeX equations directly in the WordPress editor, which is cool. Here is the single Logistic model again

Since Fluid Blue is a GPL licensed theme, here is an archive of the (minimal) changes I made.

Latexrender is useful, since it has AMS LaTeX font support, for example I can write things like , and , which is cool.

Its still a shame that in 21st century its still so hard to write decently typeset mathematics, but at least thanks to the GNU free software movement, Donald Knuth’s open source Tex, and php things at least look reasonable. It’s a shame that MathML development has been so poor and slow. Hopefully, with the adoption of UTF-8 (Unicode) as the standard encoding scheme, and the open source Unicode mathematical font project STIX about finished, there is a reasonable chance that LaTeX will be able to typeset UTF-8 Unicode fonts + html5 and finally math on the web will look as good as PDF.

After a lot of hacking, I managed to get a LaTeX plugin for Wordpress working on (mt) (gs).

which was entered by:
square bracket tex \frac{\kappa}{ 1 + \exp(\alpha \, t - \beta)}

I’m not thrilled with the rendering quality, but I can probably tweak that now that I have everything working…

First, (mt) (gs) has latex installed, but misconfigured. Tech support was unwilling to help, so I figured out how to download the new TeX user group TeXLive (strange name) unix installer, and install it as a user through ssh. Fun. But it actually worked. Then, this site http://fugato.net/2007/01/20/latex-in-wordpress/ and http://sixthform.info/steve/wordpress/ led the way for wordpress and php. Cool when it all works!!!.

Test of 2nd equation: latexrender does not appear to work with wordpress 2.7.x RC…

how about AMS math fonts: cool!

http://lizardinthesun.com/loglet/loglet3.html

is an alpha release of a javascript version of Loglet Lab, which fits S-curves of the form:

For a demo, click the buttons labeled “step 1″, “step 2″, and “step 3″ (in that order). That will load the “classic” dataset on the growth of a sunflower, and fit a single S-shaped logistic to it. The data and plots all load on one page, so you have to keep scrolling down. There are other datasets pre-loaded. The data is “live” once loaded, so you can change values and click “step 3″ (fit) with new values.

This “app” is mainly to test browser support and work on the algorithms and plotting, with the eventual goal of turning this into a google spreadsheet widget, but it is actually usable on its own.

I am using a “new” constrained regression algorithm that is inefficient (but robust…). So you really need a cutting edge browser (i.e. google chrome, webkit nightly build, or Firefox 3.1 beta with JIT enabled). I don’t use Microsoft IE, so it probably won’t work on IE. I plan on explaining the regression algorithm in a later post.

This PDF explains a bit about the logistic formulation used and how we do our plots.

I do all my numerical algorithm development in Matlab, so I needed a way to easily get results from Matlab into javascript. Here is a (a pretty cool if I don’t say so myself) matlab code that creates a JSON file from a matlab data structure.

data.x = rand(10,1);
data.y = rand(10,1)

dataname = 'data' 

fid = fopen('matlab.json','w');

numdata = max(size(fieldnames(eval(dataname))))

fdata = char(fieldnames(eval(dataname)));

fprintf(fid,'{ "%s" : {\n ',dataname)
for dataiter=1:numdata,
  fprintf(fid,' "%s" : [ \n',fdata(dataiter));
  stmp = sprintf('%s.%s',dataname,fdata(dataiter));
  for i=1:(max(length(eval(stmp))))-1,
    stmp = sprintf('%s.%s(%i)',dataname,fdata(dataiter),i);
    fprintf(fid,'%1.10e , \n',eval(stmp));
  end
  stmp2 =  sprintf('%s.%s',dataname,fdata(dataiter));;
  stmp = sprintf('%s.%s(%i)',dataname,fdata(dataiter),(max(length(eval(stmp2)))));
  fprintf(fid,'%1.10e  \n',eval(stmp));

  if dataiter == numdata ,
    fprintf(fid,'] \n');
  else
    fprintf(fid,'], \n');
  end

end

fprintf(fid,'} } \n');

And then reading it in and parsing it with jQuery is really simple.

 var x = [];
 var y = [];
 $.ajaxSetup({ async: false});
  $.getJSON("matlab.json", function (json)
            {
              x = json.data.x.slice(0);
              y = json.data.y.slice(0);
            }
            );