ALPHAPADI: FORMULAIC ALPHA DISCOVERY VIA POOL-AWARE HIERARCHICAL DISCRETE DIFFUSION
By Yanzheng Jin, Pengyang Shao, Yunshan Ma, Haowen Pan, Naixin Zhai, Chen-Hui Song, Fei Shen, Kenji Kawaguchi
Rating
1768
Battle Count: 52
Relevance
10/10
Directly addresses the core problem of alpha discovery in quantitative trading. Proposes a novel method to improve the joint predictive power and diversity of alpha pools, which is critical for portfolio construction.
Implementation Complexity
8/10
High complexity due to the combination of discrete diffusion, hierarchical masking, grammar-constrained decoding, and preference optimization. Requires careful tuning of diffusion levels, reward weights, and buffer management.
Reproducibility
4/5
The paper provides detailed algorithm pseudocode, hyperparameter settings, and links to anonymized code. It specifies the formula vocabulary and evaluation protocol clearly. However, stochastic nature of diffusion and initialization may lead to variance in results.
About this paper
Methodology: AlphaPADI. Problem types: Generative Modeling, Portfolio Optimization, Optimization, Structured Prediction.
The interactive Everscope explorer (charts, battles, favorites) loads below.