Reproduce “Randomized Signature Methods in Optimal Portfolio Selection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Erdinc Akyildirim, Matteo Gambara, Josef Teichmann, Syang Zhou
The paper empirically demonstrates that randomized signature methods, combined with a reservoir computing framework, effectively provide non-linear and non-parametric drift estimations for optimal portfolio selection. This approach consistently outperforms traditional benchmarks in maximizing Sharpe Ratio on both simulated and real-world S&P 500 data, even when accounting for transaction costs.
Key insights
- Randomized signatures, when integrated into a reservoir computing framework, provide a computationally feasible way to capture complex non-linear and path-dependent information in high-dimensional financial data.
- Accurate, non-linear drift estimation is possible even with low signal-to-noise ratios, leading to superior portfolio performance when combined with robust covariance estimation and practical constraints.
- Careful management of transaction costs through post-processing strategies is crucial for the real-world applicability and sustained out-performance of advanced portfolio optimization techniques.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.