Signature-Informed Transformer for Asset Allocation

Modern deep learning for asset allocation typically separates forecasting from optimization. We argue this creates a fundamental mismatch where minimizing prediction errors fails to yield robust portfolios. We propose the Signature Informed Transformer to address this by unifying feature extraction and decision making into a single policy. Our model employs path signatures to encode complex path dependencies and introduces a specialized attention mechanism that targets geometric asset relationships. By directly minimizing the Conditional Value at Risk we ensure the training objective aligns with financial goals. We prove that our attention module rigorously amplifies signature derived signals. Experiments across diverse equity universes show our approach significantly outperforms both traditional strategies and advanced forecasting baselines. The code is available at: https://github.com/Yoontae6719/Signature-Informed-Transformer-For-Asset-Allocation

This study presents a new deep learning method for building investment portfolios. Instead of first predicting asset returns and then optimizing the portfolio separately, the model learns both steps together. It uses path signatures and attention mechanisms to better capture market patterns and relationships between assets. By directly focusing on reducing downside risk, the method creates more robust portfolios and performs better than existing approaches in experiments.