Relation-Aware Transformer for Portfolio Policy Learning

Authors: Ke Xu, Yifan Zhang, Deheng Ye, Peilin Zhao, Mingkui Tan

IJCAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on real-world crypto-currency and stock datasets verify the state-of-the-art performance of RAT.
Researcher Affiliation Collaboration Ke Xu1,2 , Yifan Zhang1,2 , Deheng Ye3 , Peilin Zhao3 , Mingkui Tan1 1South China University of Technology, Guangzhou, China 2Pazhou Lab, Guangzhou, China 3Tencent AI Lab, Shenzhen, China
Pseudocode No The paper describes the architecture and algorithms in prose and mathematical equations but does not include any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes 1The source code is available: https://github.com/Ivsxk/RAT.
Open Datasets Yes All crypto-currency datasets are accessed with Poloniex2, where data selection is based on the method in [Jiang et al., 2017]. We also evaluate our methods on the S&P500 stock dataset obtained from Kaggle3.
Dataset Splits No The paper provides statistics for training and test datasets in Table 1, but it does not explicitly mention or describe a separate validation dataset split with specific percentages or sample counts.
Hardware Specification Yes In the training process, we adopt Adam optimizer on a single NVIDIA Tesla P40 GPU.
Software Dependencies No The paper states that "RAT is implemented via pytorch" but does not provide specific version numbers for PyTorch or any other software dependencies.
Experiment Setup Yes The number of attention heads is set to H=2, and the dimension of the feature space is set to df=12. In the training process, we adopt Adam optimizer on a single NVIDIA Tesla P40 GPU. The training step is 80000 for crypto-currency data and 20000 for stock data, where the batch size is 128. We set learning rate to 10 4 and weight decay of l2 regularizer to 10 7. The transaction cost rate is 0.25%. The temporal length of the local context is set to l=5, while the length of the price series is k=30. For all RL based methods, results are averaged over 5 runs with random initialization seeds.