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 Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..

Towards Enhancing Time Series Contrastive Learning: A Dynamic Bad Pair Mining Approach

Authors: Xiang Lan, Hanshu Yan, Shenda Hong, Mengling Feng

ICLR 2024 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through extensive experiments conducted on four large-scale, real-world time series datasets, we demonstrate DBPM s efficacy in mitigating the adverse effects of bad positive pairs.
Researcher Affiliation Collaboration 1Saw Swee Hock School of Public Health & Institute of Data Science, National University of Singapore 2Byte Dance, 3National Institute of Health Data Science, Peking University
Pseudocode Yes A.1 DBPM ALGORITHM Algorithm 1: Dynamic Bad Pair Mining
Open Source Code Yes Corresponding authors Codes are available at Git Hub
Open Datasets Yes Our model is evaluated on four real-world benchmark time series datasets: PTB-XL (Wagner et al., 2020)... HAR (Anguita et al., 2013)... Sleep-EDF (Goldberger et al., 2000)... Epilepsy (Andrzejak et al., 2001)...
Dataset Splits Yes Table 3: Summarization of PTB-XL. Task Diagnostic Classification Train 13,715 Val 3,429 Test 4,286 and Table 4: Summarization of HAR, Epilepsy, and Sleep-EDF. Dataset HAR Train 5,881 Val 1,471 Test 2,947
Hardware Specification Yes Experiments are conducted using Py Torch 1.11.0 (Paszke et al., 2019) on a NVIDIA A100 GPU.
Software Dependencies Yes Experiments are conducted using Py Torch 1.11.0 (Paszke et al., 2019) on a NVIDIA A100 GPU.
Experiment Setup Yes The Adam optimizer (Kingma & Ba, 2015) with a fixed learning rate of 0.001 is used to optimize the linear classifier for all datasets. The temperature t is set to 0.2 for the contrastive loss defined in Eq.1.