Unsupervised Feature Learning from Time Series

Authors: Qin Zhang, Jia Wu, Hong Yang, Yingjie Tian, Chengqi Zhang

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments show that USLM outperforms search-based algorithms on real-world time series data.
Researcher Affiliation Collaboration Quantum Computation & Intelligent Systems Centre, University of Technology Sydney, Australia Research Center on Fictitious Economy & Data Science, Chinese Academy of Sciences, Beijing, China. Key Lab of Big Data Mining & Knowledge Management, Chinese Academy of Sciences, Beijing, China. Math Works, Beijing, China
Pseudocode Yes Algorithm 1. Unsupervised Shapelet Learning Algorithm (USLA) 1: Input: Time series T with c classes Length and number of shapelets: lmin, r, k Number of internal iterations imax Learning rate and Parameters λ1, λ2, λ3 and , σ 2: Output: Shapelets S and class labels Y 3: Initialize: S0, W0, Y0 4: While Not convergent do 5: Calculate: Xt(T, St| ), LGt(T, St| , σ) 6: and Ht(St| ) based on Eqs. (2), (4), and (6); 7: update Wt+1, Yt+1: 8: Yt+1 λ2WT t Xt(LGt + λ2I) 1 9: Wt+1 (λ2Xt XT t + λ3I) 1(λ2Xt YT t+1). 10: update St+1: 11: for i = 1, . . . , imax do 12: Si+1 Si r Si 13: r Si = @F(Si|Xt+1,Yt+1) @S is from Eq. (17) 14: end for 15: St+1 = Simax+1 16: t t + 1 17: end while 18: Output: S = St; Y = Yt; W = Wt.
Open Source Code Yes The Matlab source codes and data are available online1. 1https://github.com/Blind Review/shapelet
Open Datasets Yes We use seven time series benchmark datasets download from the UCR time series archive [Chen et al., 2015] [Cetin et al., 2015]. The UCR time series classification archive, www.cs.ucr.edu/ eamonn/time series data/. 2015.
Dataset Splits No Table 1 lists "Train/Test" splits for the datasets (e.g., CBF 30/900(930)), but there is no explicit mention or description of a validation dataset split.
Hardware Specification Yes All experiments are conducted on a Windows 8 machine with 3.00GHz CPU and 8GB memory.
Software Dependencies No The paper mentions "Matlab source codes" but does not specify the version of Matlab or any other software dependencies with version numbers.
Experiment Setup Yes The remaining parameters are fixed as follows, λ1 = λ2 = λ3 = λ4 = 1, σ = 1, Imax = 50, = 0.01 and the length of the shapelets is set to 10.