Open Anomalous Trajectory Recognition via Probabilistic Metric Learning
Authors: Qiang Gao, Xiaohan Wang, Chaoran Liu, Goce Trajcevski, Li Huang, Fan Zhou
IJCAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on two large-scale trajectory datasets demonstrate the superiority of ATROM in addressing both known and unknown anomalous patterns. |
| Researcher Affiliation | Academia | 1Southwestern University of Finance and Economics, Chengdu, China, 611130 2University of Electronic Science and Technology of China, Chengdu, China, 610054 3Iowa State University, Iowa, USA |
| Pseudocode | No | The paper describes its methodology and components but does not include a clearly labeled 'Pseudocode' or 'Algorithm' block. |
| Open Source Code | Yes | For reproducibility, the source codes are available at https://github.com/ypeggy/ATROM. |
| Open Datasets | Yes | We conduct our experiments on two real-world taxi trajectory datasets. The first taxi dataset [Liu et al., 2020] is collected from 442 taxis operating in the city of Porto during Jan 07 2013 to Jun 30 2014. ... The second dataset is collected from Di Di Chuxing1, containing a large number of taxi traces generated from the city of Chengdu in Aug 2014. ... 1http://outreach.didichuxing.com/research/opendata/ |
| Dataset Splits | No | For the training and testing setups, we use 90% of the trajectories as the training set and the rest as the testing set. The paper does not explicitly mention a separate validation set split. |
| Hardware Specification | Yes | We implemented ATROM and all the baselines in Python using the Py Torch library, accelerated by the NVIDIA Tesla A100. |
| Software Dependencies | No | The paper mentions 'Python using the Py Torch library' but does not specify version numbers for either, which is required for reproducibility of software dependencies. |
| Experiment Setup | Yes | In the implementation, we set d to 128, J is 10, the size of hidden state is 256, and the learning rate is initialized as 0.001. We use Adam as the optimization algorithm. |