End-to-End Python implementation of bi-level optimization for financial time-series synthesis from "History Is Not Enough" by Xia et al. (2026). Implements cointegration-aware data augmentation, curriculum learning, and meta-learned augmentation policies to immunize quantitative models against concept drift and market non-stationarity.
python reinforcement-learning deep-learning pytorch scientific-computing neural-networks quantitative-finance bilevel-optimization algorithmic-trading data-augmentation cointegration concept-drift time-series-analysis market-microstructure meta-learning financial-econometrics curriculum-learning time-series-forecasting financial-machine-learning non-stationarity
-
Updated
Jan 19, 2026 - Jupyter Notebook