Hierarchical Multi-Task Learning with Liquidity-Aware Signals for Stock Forecasting
By Hengyi Yang, Sida Lin, Yiyan Qi, Yankai Chen, Haohan Zhang, Xianhua Peng, Jian Guo
Rating
1500
Battle Count: 0
Relevance
9/10
Highly relevant as it directly addresses the gap between predictive accuracy and executable portfolio performance by integrating liquidity constraints and transaction costs into the learning and optimization pipeline.
Implementation Complexity
8/10
Complex architecture involving hierarchical attention, Mixture-of-Experts with cross-task gating, and a custom portfolio optimization layer. Requires careful tuning of task weights and gating parameters.
Reproducibility
4/5
Code is available via an anonymized repository. The paper uses public datasets (CSI300/CSI500 via Qlib) and provides detailed hyperparameters and implementation specifics.
About this paper
Methodology: LiMT (Hierarchical Multi-Task Learning with Liquidity-Aware Signals). Problem types: Time Series Forecasting, Regression, Multi-task Learning, Portfolio Optimization, Risk Management.
The interactive Everscope explorer (charts, battles, favorites) loads below.