Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

Classification of Time Sequences using Graphs of Temporal Constraints

Authors: Mathieu Guillame-Bert, Artur Dubrawski

JMLR 2017 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We explain the proposed algorithms and evaluate their performance against a diverse collection of 59 benchmark data sets. In these experiments, our algorithms come across as highly competitive and in most cases closely match or outperform state-of-the-art alternatives in terms of the computational speed while dominating in terms of the accuracy of classification of time sequences.
Researcher Affiliation Academia Mathieu Guillame-Bert EMAIL Artur Dubrawski EMAIL Auton Lab, The Robotics Institute, School of Computer Science Carnegie Mellon University, Pittsburgh, United States
Pseudocode Yes Pseudocode in Algorithm 1 provides details. Algorithm 1: Extraction of a GTC decision forest from data. ... Algorithm 2: Extraction of a GTC set. ... Algorithm 3: Finding the optimal threshold value α for a new scalar test. ... Algorithm 4: Building matrix P.
Open Source Code No The synthetic data set as well as a Python script used to generate it are available at mathieu.guillame-bert.com/dataset. (This link is for the dataset and a script to generate it, not the source code for the proposed GTC-DF or GTC-Set algorithms.)
Open Datasets Yes The synthetic data set as well as a Python script used to generate it are available at mathieu.guillame-bert.com/dataset. ... The University of California at Riverside (UCR) benchmark data is a collection of 41 diverse data sets representing time series classification problems of varying complexity. ... These data are available at http://www.cs.ucr.edu/ eamonn/time series data. ... The internal bleeding detection data set is available at mathieu.guillame-bert.com [Guillame-Bert].
Dataset Splits Yes All reported results have been computed using the same 10-fold partitioning of data. In the case of the rotated UCR data sets, the 10-fold cross-validation is repeated for each of the 10 rotations of each rotated data set (e.g. each reported error rate is computed from 100 train and test experiments). In Bleeding Detection Data Set, we ensure that the complete record for each animal is either entirely used for training or for testing in each cross-validation iteration.
Hardware Specification Yes Experiments have been performed on a 3.4GHz i7 8-core processor with 16GB of main memory.
Software Dependencies No All considered algorithms have been implemented (or re-implemented) in C++ to ensure fairness of comparison. (The paper mentions C++ as the implementation language but does not specify version numbers for C++ or any libraries/compilers used.)
Experiment Setup Yes Algorithm parameters have not been tuned and were assigned to their meaningful default settings (except for k in ED+k NNCV ), as follows. RF: max Num Trees(30), min Num Observations(5), max Depth(10). LCSS+NN : e(0.05 standard deviation). EDR+NN : e(0.25 standard deviation). LCSSC+NN : The maximum window is set to 25% of the data set length.