Towards a Minimum Description Length Based Stopping Criterion for Semi-Supervised Time Series Classification

Nurjahan Begum, Bing Hu, Thanawin Rakthanmanon, Eamonn Keogh

================================================================================================================================================

 

Supporting Material

 

1.      Slide* showing all figures in better resolution

2.      Matlab source code (SSL*, MDL*)

 

 

MIT-BIH Supraventricular Arrhythmia Database

1.      A heartbeat dataset from the PhysioBank ATM. (Dataset: svdb, Record: 801, Signal: ECG1)

2.      Mirror* of the data

 

 

St. Petersburg Arrhythmia Database

1.      A heartbeat dataset from the PhysioBank ATM. (Dataset: incartdb, Record: I70, Signal: II)

2.      Mirror* of the data

 

 

Sudden Cardiac Death Holter Database

 

1.      A heartbeat dataset from the PhysioBank ATM. (Dataset: sddb, Record:52, Signal: ECG(1st))

2.      Mirror* of the data

 

 

Swedish Leaf Dataset

 

Ulmus carpinifolia

 

Acer platanoides

 

Ulmus

 

Quercus robur

 

Alnus incana

 

Tilia

 

Salix fragilis

 

Populus tremula

 

Corylus avellana

 

Sorbus aucuparia

 

Prunus padus

 

    Tilia

 

Populus

 

Sorbus hybrida

 

Fagus silvatica

 

 

*Oskar J. O. Soderkvist. Computer Vision Classification of Leaves from Swedish Trees, Master thesis, Linkoping University, Sweden, 2001.

1.      Contains leaf images from 15 tree classes. The images are converted to time series.

2.      Mirror* of the data

3.      Seed instance chosen for experiment: Fagus silvatica

 

 

Fish Dataset

*image ref: http://jmotif.googlecode.com/svn/trunk/RCode/fish/fish.png

1.      Contains fish images and contains 7 classes. The images are converted to time series.

2.      Mirror* of the data

3.      Seed instance chosen for experiment: class 2

 

 

Face_all Dataset

*image ref: http://jmotif.googlecode.com/svn/trunk/RCode/faceAll/all_classes.png

1.      Contains face shapes and is of 14 classes. The images are converted to time series.

2.      Mirror* of the data

3.      Seed instance chosen for experiment: class 2

 

 

Extra Experiments

 

Extra experiments expressing the generality of our algorithm can be found from here*.

 

 

 

 

 

* All files are password protected, for the password please send an email to: nbegu001 "at" cs "dot" ucr "dot" edu