pyts: A Python Package for Time Series Classification

Authors: Johann Faouzi, Hicham Janati

JMLR 2020 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Table 2: Accuracy scores on the test set for the BOSS transformer followed by a one-nearest neighbor classifier with the BOSS metric on several data sets.
Researcher Affiliation Academia Johann Faouzi EMAIL Aramis Lab, INRIA Paris Brain and Spine Institute 75013 Paris France; Hicham Janati EMAIL Parietal team, INRIA Saclay Neurospin, Bˆat 145 CEA Saclay 91191 Gif sur Yvette France
Pseudocode Yes Listing 1: Code snippet illustrating pyts s intuitive API on a classification example.
Open Source Code Yes pyts is an open-source Python package for time series classification. ... Source code and documentation can be downloaded from https://github.com/johannfaouzi/pyts.
Open Datasets Yes X_train , X_test , y_train , y_test = load_gunpoint(return_X_y=True); Table 2: Accuracy scores on the test set for the BOSS transformer followed by a one-nearest neighbor classifier with the BOSS metric on several data sets. Adiac ECG200 Gun Point Middle Phalanx TW Plane (Sch afer, 2015)
Dataset Splits Yes X_train , X_test , y_train , y_test = load_gunpoint(return_X_y=True)
Hardware Specification No No specific hardware details (like GPU/CPU models, memory) are mentioned in the paper.
Software Dependencies Yes Packages (versions): pyts (0.10.0), sktime (0.3.1), tslearn (0.2.5), seglearn (1.1.0), tsfresh (0.13.0), cesium (0.9.10).
Experiment Setup Yes clf = BOSSVS(window_size =28)