Reproduce “FlashFolio: A GPU-Accelerated Solver for Portfolio Optimization” 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
Yilun Jiang, Haihao Lu, Zedong Peng, Jinwen Yang
Researchers developed FlashFolio, a GPU-accelerated solver for large-scale single- and multi-period portfolio optimization, capable of handling factor-based risk and nonlinear transaction costs. The solver achieved speedups of 2.4x to 12.9x for single-period models and 15.8x to 48x for multi-period models, demonstrating superior numerical robustness compared to a CPU-based conic solver.
Key insights
- Modern GPUs offer massive parallelism and high memory bandwidth, presenting an opportunity to significantly accelerate numerical operations central to large-scale convex optimization problems in finance.
- Tailoring solver algorithms to efficiently map complex portfolio optimization problem structures onto GPU architectures can yield substantial computational efficiency gains.
- Overcoming computational bottlenecks via GPU acceleration improves the practicality of deploying sophisticated multi-period models in latency-sensitive real-time trading and risk management systems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.