DiffTraj: Generating GPS Trajectory with Diffusion Probabilistic Model

Authors: Yuanshao Zhu, Yongchao Ye, Shiyao Zhang, Xiangyu Zhao, James Yu

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

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct extensive experiments on two real-world trajectory datasets to show the superior performance of the proposed Diff Traj in GPS trajectory generation.
Researcher Affiliation Academia Yuanshao Zhu1,2, Yongchao Ye1, Shiyao Zhang1, Xiangyu Zhao2, , James J.Q. Yu3, 1 Southern University of Science and Technology 2 City University of Hong Kong 3University of York
Pseudocode Yes The training and sampling phase of the proposed framework is summarized in Algorithm 1 and Algorithm 2, respectively.
Open Source Code Yes The code for the implementation of Diff Traj is available for reproducibility2. 2https://github.com/Yasoz/Diff Traj
Open Datasets Yes These datasets can be downloaded at https://outreach.didichuxing.com/
Dataset Splits No The paper mentions using real-world datasets and evaluating performance but does not specify the explicit training, validation, and test splits (e.g., percentages or counts) for the primary Diff Traj model's development.
Hardware Specification Yes All experiments are implemented in Py Torch and conducted on a single NVIDIA A100 40GB GPU.
Software Dependencies No The paper mentions that experiments are implemented in 'Py Torch' but does not specify a version number for PyTorch or any other software dependencies.
Experiment Setup Yes For the proposed Diff Traj framework, we summarize the adopted hyperparameters in Table 3. ... Table 3: Hyperparameters setting for Diff Traj. Diffusion Steps 500, Skip steps 5, Guidance scale 3, β (linear) 0.0001, Batch size 1024, Sampling blocks 4, Resnet blocks 2, Input Length 200