DTW Step performance in Trajectory

Today I've hacked in the DTW step implementation in the Trajectory. Take a look on the impact of the step choice on the Euclidean distance after alignment and plots rendering.

Symmetric Pattern 0


Asymmetric Pattern 05

Step functions performance illustrations

A simple example demonstrating the performance of different step function following "Dynamic programming algorithm optimization for spoken word recognition"
by: H Sakoe, S Chiba and using matrix from T. Giorgino (2007). "dtw: Dynamic Time Warping algorithms in R 1.12-3"

The distance matrix.


The step functions performance, red - the result of DTW alignment, black - the true path of zeroes.

Trajectory is up and running within the ProjectBrowser

I've fixed CSS styles and did some code changes within the GoogleChart package (due to the precision lost while converting from doubles I wasn't able to see normal and DTW plots at all). The trajectory runs on my workstation, will see today how it deals with bigger datasets. My next step would be to implement constraints, step patterns and open end alignment classes plugging them into UI.

Who broke the build!?

Last week I've cleaned up the Trajectory code and finally merged my branch with the trunk. I've broke the build. I'm sorry. Thank you Philip for prompt response and help fixing it.

Among other things, since I've forked it long ago, I had some troubles with code layout and Trajectory page is still messed up since the div's name changed within HTML templates. Next week will fix CSS's and HTML's for the trajectory. Good news that it performs as it suppose (at least on my workstation).

Started organizing my literature review using CiteULike.

And after all I'm coding scanner and parser parts for the compiler :P...

Progress report

I've finished modifying code architecture for the DTW and improved data normalization for the time series. The original Telemetry and GoogleChart modules code which I borrowed for modification appeared to have three or four validation and format routines which were designed the way which prevented normalized data to be rendered properly. I've changed/commented out validations and formatting but not sure if it was right decision, I feel like I need a little more time to see what was behind this formatting and will post results later.

The most current plots: