Paper Released on Det-MASO Multi-Agent Factor Generation Framework
Get up to speed
Tiantian Wu and Peng Yang released a paper on the Det-MASO multi-agent sequence optimization framework on arXiv. Addressing the limitations of existing single-agent factor generation in constrained space exploration and the tendency of multi-agent systems to output homogeneous factors, they propose introducing a differentiable log determinant penalty for the inter-agent correlation matrix into the joint objective. The computational cost grows linearly with the number of assets, suppressing agent homogenization during training. Backtests in the paper show that the framework improves the information coefficient of corresponding factors by 1.1%, increases annualized returns by 6.4%, and raises the Sharpe ratio by 0.33.
AI-generated from reports · 2 hr agoUpdated
Coverage timeline
Follow the coverage to understand different sides of the event.
- arXiv q-finDet-MASO: Diversity-Regularized Multi-Agent Sequence Optimization for Factor Generation
To address the limitations of existing single-agent factor generation methods in exploring the factor space and the tendency of multi-agent methods to produce homogeneous factors, this study proposes the Det-MASO multi-agent sequence optimization framework. The framework introduces a differentiable log-determinant penalty on the inter-agent correlation matrix in the joint objective, with computational cost growing linearly with the number of assets, which suppresses agent homogenization during training. Backtesting shows that it improves the information coefficient by 1.1%, annualized return by 6.4%, and Sharpe ratio by 0.33.
Trend of this event
Not enough consecutive observations yet; trend is not drawn.