Unsupervised Path Representation Learning with Curriculum Negative Sampling
Authors: Sean Bin Yang, Chenjuan Guo, Jilin Hu, Jian Tang, Bin Yang
IJCAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on two downstream tasks, ranking score estimation and travel time estimation, using two road network datasets suggest that PIM significantly outperforms other unsupervised methods and is also able to be used as a pre-training method to enhance supervised path representation learning. |
| Researcher Affiliation | Collaboration | 1Department of Computer Science, Aalborg University, Denmark 2Mila-Quebec AI Institute 3HEC Montreal, Canada 4CIFAR AI Research Chair {seany, cguo, hujilin, byang}@cs.aau.dk, jian.tang@hec.ca |
| Pseudocode | No | The paper describes the proposed methods and uses figures to illustrate concepts (e.g., Figure 2 for Curriculum Negative Sampling), but it does not include any formal pseudocode blocks or algorithm listings. |
| Open Source Code | Yes | Code is available at https://github.com/Sean-Bin-Yang/Path-Info Max.git. |
| Open Datasets | No | The paper states: 'We obtain two road network graphs from Open Street Map.' and 'We also obtain two substantial GPS trajectory data sets on the two road networks.' While it names the sources (Open Street Map, GPS trajectory data), it does not provide a direct link, DOI, specific repository name, or formal citation for the prepared datasets used in the experiments, nor does it explicitly state their public availability beyond their raw sources. |
| Dataset Splits | Yes | We randomly choose 85%, 10%, and 5% of the paths as the training, validation, and test sets. |
| Hardware Specification | Yes | We conduct experiments on Ubuntu 18.04.5 LTS, with 40 Intel(R) Xeon(R) Gold 5215 CPUs @ 2.50GHz and four Quadro RTX 8000 GPU cards. |
| Software Dependencies | Yes | All algorithms are implemented in Pytorch 1.7.1. |
| Experiment Setup | Yes | We use an LSTM as the path encoder. We use node2vec [Grover and Leskovec, 2016], an unsupervised node representation learning method, to obtain a 128 dimensional node feature vector for each node, i.e., D = 128. We set the path representation size D = 128. In the curriculum negative sampling, for each input path, we generate four negative paths... We use Adam [Kingma and Ba, 2015] for optimization with learning rate of 0.001. |