My R package for finding time series similarity based on learned patterns is on CRAN

Finally, I could have finished implementing the R package for our recent approach "Learned Pattern Similarity (LPS)" for time series representation and similarity. An earlier version was made available on this web site when we submitted our PAMI paper but the Comprehensive R Archive Network (CRAN) has some requirements (such as manuals, descriptions and etc.) before releasing it to open-source community.

The first version is now available at http://cran.r-project.org/web/packages/LPStimeSeries/index.html. I was planning to name the package as "LPS" but unfortunately there is one package named as LPS that was released in December, 2013 (after our paper submission). Hence, I have named the package as LPStimeSeries which stands for "Learned Pattern Similarity and Representation for Time Series".

The implementation is little different than the one in the package available here. I have many things in to-do list to work on and I will get started soon. 

If you are planning to compare LPS to a time series similarity/representation approach, please use the package on this website (Update: You can use the R package made available by CRAN). Also, I made a simple MATLAB implementation available. You may want to try that too but the regression tree implementation is different than my implementation in the R package. Hence, some parameters may not exist for MATLAB. For example, in R we have a depth parameter where you can restrict the number of instances in a node to control depth in MATLAB. 

Let me know if you have any questions.

Copyright © 2014 mustafa gokce baydogan

LinkedIn
Twitter
last.fm