Skip to content

Latest commit

 

History

History
81 lines (72 loc) · 2.71 KB

README.md

File metadata and controls

81 lines (72 loc) · 2.71 KB
Raw

Preamble

This code is the one used to generate results presented in the paper Efficient Temporal Kernels between Feature Sets for Time Series Classification. When using this code, please cite:

@inproceedings{tavenard:halshs-01561461,
  TITLE = {{Efficient Temporal Kernels between Feature Sets for Time Series Classification}},
  AUTHOR = {Tavenard, Romain and Malinowski, Simon and Chapel, Laetitia and Bailly, Adeline and Sanchez, Heider and Bustos, Benjamin},
  URL = {https://halshs.archives-ouvertes.fr/halshs-01561461},
  BOOKTITLE = {{European Conference on Machine Learning and Principles and Practice of Knowledge Discovery}},
  ADDRESS = {Skopje, Macedonia},
  YEAR = {2017},
  MONTH = Sep,
  KEYWORDS = {Time series classification},
  PDF = {https://halshs.archives-ouvertes.fr/halshs-01561461/file/paper.pdf},
  HAL_ID = {halshs-01561461},
  HAL_VERSION = {v1},
}

Supplementary material

Supplementary material for this paper is available here.

Requirements

For this code to run properly, the following python packages should be installed:

numpy  
scipy  
sklearn

Also, if one wants to run experiments on the UCR dataset, she should download it from here. Then, using the software from https://github.com/a-bailly/dbotsw, she should generate time-sensitive features (cf. get_feature_vectors) into the folder datasets/ucr_t with class information inside datasets/ucr_classes.

Running

Baseline (BoW)

To run the baseline on dataset FISH with $k=1024$, run:

SOURCEDIR=/path/to/the/base/dir/of/the/project/
WORKINGDIR=${SOURCEDIR}/
EXECUTABLE=${SOURCEDIR}/ucr_bow.py
export PYTHONPATH="${PYTHONPATH}:${SOURCEDIR}"
cd ${WORKINGDIR}
python ${EXECUTABLE} FISH 1024

SQFD-KM

To run the SQFD-KM method on dataset FISH with $k=256$, run:

SOURCEDIR=/path/to/the/base/dir/of/the/project/
WORKINGDIR=${SOURCEDIR}/
EXECUTABLE=${SOURCEDIR}/ucr_sqfd_km.py
export PYTHONPATH="${PYTHONPATH}:${SOURCEDIR}"
cd ${WORKINGDIR}
python ${EXECUTABLE} FISH 256

SQFD-Nystroem

To run the SQFD-Nystroem method on dataset FISH with $D=1024$, run:

SOURCEDIR=/path/to/the/base/dir/of/the/project/
WORKINGDIR=${SOURCEDIR}/
EXECUTABLE=${SOURCEDIR}/ucr_sqfd_nystroem.py
export PYTHONPATH="${PYTHONPATH}:${SOURCEDIR}"
cd ${WORKINGDIR}
python ${EXECUTABLE} FISH 1024

SQFD-Fourier

To run the SQFD-Fourier method on dataset FISH with $D=1024$, run:

SOURCEDIR=/path/to/the/base/dir/of/the/project/
WORKINGDIR=${SOURCEDIR}/
EXECUTABLE=${SOURCEDIR}/ucr_sqfd_fourier.py
export PYTHONPATH="${PYTHONPATH}:${SOURCEDIR}"
cd ${WORKINGDIR}
python ${EXECUTABLE} FISH 1024