.comment-link {margin-left:.6em;}

Blog site moved to new location...

The Pile

Using a feed:
Using e-mail:: Your e-mail

Wednesday, July 28, 2004

Mathematica Eye Candy

I have been working on a little project that involves data mining some personal history. One of the things I have been trying is using principal component analysis to reduce the dimensionality of my data to something I can get an intuitive feel for, so that I can try to ficure out what are the best automated methods for pattern recognition. This image is an example reduced from sixteen variables to three. It's impressive what you can do in Mathematica.

posted by Eamonn | 6:34 PM | 0 comments

Tuesday, July 27, 2004

Dinner at Scott Seafood, Palo Alto

We had a nice half-bottle of Acacia 2001 pinot noir, Napa, Carneros

I had a nice salad, a tasty rare Ahi tuna, with an Illy espresso to finish. P was not so lucky, with a salad that she found overdressed, and a fish dish that both she and I found overdressed.

posted by Eamonn | 9:49 PM | 0 comments

Applet Demos of Wolfram Cellular Automota

Rudy Rucker and his class have done a good job of providing
applet versions of some of the cellular automota from Stephen Wolfram's amazing book book "A New Kind of Science".

I read the book when it first came out. It is nice to be reminded again at how amazing it all is.

posted by Eamonn | 9:44 AM | 0 comments

Monday, July 26, 2004

SDForum

Some interesting technical events sponsored by the SDForum.

posted by Eamonn | 3:31 PM | 0 comments

Saturday, July 24, 2004

Clustering Algorithms and Dimensional Reduction

I am giving my self a crash course on some data mining techniques for a project I am working on. Here are some things I found useful.

In Matteo Matteucci's site at Politechnico di Milano is a nice little introductory tutorial on clustering algorthms, complete with interactive demos. A similar page is Tariq Rashid's University of Bristol page

Also François Labelle at McGill has a nice overview of reducing the dimensionality of multivariate data using Principal Component Analysis, also with interactive demos which give a nice intuitive feel for the technique. Mathematica supports principal component analysis, so given a data matrix with the each observation in a row, and each column a dimension I found could do the following to get a nice two dimensional view of the multi-dimenstional data:


<<Statistics`MultiDescriptiveStatistics`

rotated = PrincipalComponents[Transpose[data]];

rotated2d = Table[ {rotated〚i,1〛, rotated〚i,2〛}, {i,1,n}];

ListPlot[rotated2d]

posted by Eamonn | 12:42 PM | 0 comments

Tuesday, July 20, 2004

Whale watching

Just back from an amazing whale-watching boat trip on the straits between San Juan Island, Washington State and Vancouver Island, British Columbia. We were lucky in that all of the local pods of Orca had come together to socialize and hunt for the salmon sheltering from the strong ebb tide currents in coves along the coast. We saw lots of the orca swimming along in family groups, splashing about as they herded the salmon, and popping their heads out of the water to keep an eye on the boats.

posted by Eamonn | 4:53 PM | 0 comments

Tuesday, July 13, 2004

Reconstructing grammars from sequences

Interesting. I thingk I might be able to use this in my "personal data mining" experiments.

posted by Eamonn | 9:29 PM | 0 comments

Tuesday, July 06, 2004

Memory footprint of Java Objects.

A C/C++ programmer, used to using sizeof, might be suprised how hard it is to figure out the memory footprint of Java object. Of course relying on memory sizes is very bad for portability, but sometimes when tuning applications you do need this information.

This article walks through some measurments that reveal:

Object	Size (bytes)	Overhead (bytes)	Overhead (percent)
Object	8
Integer	16	12	300%
Long	16	8	100 %
int[n]	16 + 4*n	16
char[n]	16 + 2*n	16
String, length n	40 + 2*n	40

posted by Eamonn | 11:54 AM | 0 comments

Sunday, July 04, 2004

Blogging from my phone.

Just testing to see if I can post to this blog from my phone.

posted by Eamonn | 11:22 PM | 0 comments

Glimpses of the surface of titan.

Finally we are starting to be able to see some detail of Titan's surface through the haze of its atmosphere.

http://photojournal.jpl.nasa.gov/catalog/PIA06407

posted by Eamonn | 10:27 PM | 0 comments