Joint Representation Learning for Multi-Modal Transportation Recommendation
Authors: Hao Liu, Ting Li, Renjun Hu, Yanjie Fu, Jingjing Gu, Hui Xiong1036-1043
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments Using large-scale real-life datasets, we present an extensive experimental study to evaluate: (1) the overall performance of Trans2Vec, (2) the parameter sensitivity, (3) the transport mode relevance and (4) the robustness of our approach. |
| Researcher Affiliation | Collaboration | Hao Liu,1 Ting Li,2 Renjun Hu,3 Yanjie Fu,4 Jingjing Gu,5 Hui Xiong1 1The Business Intelligence Lab, Baidu Research, National Engineering Laboratory of Deep Learning Technology and Application, Beijing, China, 2National University of Defense Technology, Changsha, China 3SKLSDE Lab, Beihang University, Beijing, China, 4Missouri University of Science and Technology, Missouri, USA 5Nanjing University of Aeronautics and Astronautics, Nanjing, China |
| Pseudocode | Yes | Algorithm 1: Joint learning algorithm of Trans2Vec |
| Open Source Code | No | The paper states that Trans2Vec 'has been deployed in a map and navigation App' but does not provide any information or links regarding the public availability of its source code. |
| Open Datasets | No | We choose two datasets BEIJING and SHANGHAI to test our approach. Both of them are produced based on the map queries and user feedbacks on the Baidu Map, in the corresponding cities from April 1, 2018 till August 20, 2018. |
| Dataset Splits | No | We use the data from April through July for training, i.e., learning embeddings, counting historical numbers of rides and training models, and the rest data for testing. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU or CPU models, or server configurations used for running the experiments. |
| Software Dependencies | No | The paper does not provide specific version numbers for software dependencies or libraries used in the experiments. |
| Experiment Setup | Yes | We use the recommended parameters for all baselines, and fixed learning rate α = 0.5 and 0.3 for BEIJING and SHANGHAI, respectively, number d of dimensions to 64, number K of relevance neighbors to 5, and regularizing parameters β1, β2 and γ to 0.1, 0.3 and 0.5, respectively, by default for our BTrans2Vec and Trans2Vec (we will report the sensitivity analysis later). |