Contrastive Pre-training with Adversarial Perturbations for Check-In Sequence Representation Learning

Authors: Letian Gong, Youfang Lin, Shengnan Guo, Yan Lin, Tianyi Wang, Erwen Zheng, Zeyu Zhou, Huaiyu Wan

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the effectiveness and versatility of CACSR on two kinds of downstream tasks using three real-world datasets. The results show that our model outperforms both the state-of-the-art pre-training methods and the end-to-end models.
Researcher Affiliation Academia 1School of Computer and Information Technology, Beijing Jiaotong University, Beijing, China 2Beijing Key Laboratory of Traffic Data Analysis and Mining, Beijing, China
Pseudocode No The paper includes figures illustrating the model architecture and workflow, but it does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes The code has been released at: https://github.com/LetianGong/CACSR
Open Datasets No The paper mentions using "three real-world datasets derived from the raw Gowalla, and Foursquare of New York City (NYC) and Jakarta (JKT) check-in data." However, it does not provide specific links, DOIs, or citations with author/year information for direct access to these derived datasets or the raw data from which they were obtained.
Dataset Splits Yes We split all datasets at ratio 6 : 2 : 2 into training sets, validation sets, and test sets by the samples.
Hardware Specification Yes All trials have been conducted on Intel Xeon E5-2620 CPUs and NVIDIA RTX A5000 GPUs.
Software Dependencies No The paper mentions using "Py Torch" but does not specify a version number (e.g., PyTorch 1.9), nor does it list any other software components with their versions.
Experiment Setup Yes As for the parameter settings, we set all embedding sizes of all models to 256. The number layer of Bi-LSTM in CACSR model is set to 3, the hidden state size is set to 512, σ = 0.1, the scale factor η = 1, ϵ = 1, α = 0.8, β = 0.5, and τ = 4. The CACSR is pre-trained for 100 epochs on the training sets with the early-stopping mechanism of 5 patience.