End-to-End Learning of Financial Networks for Futures Momentum

Summary

The document presents L2GMOM, a machine-learning framework for network momentum. Network momentum uses relationships among assets to help predict future returns. The proposed framework learns the financial network and trading signals together, instead of constructing a network first and optimizing a portfolio as a separate step. Its neural network architecture is derived from algorithm unrolling and is described as interpretable. The framework can use different portfolio-performance objectives, including the negative Sharpe ratio.

Backtests on 64 continuous futures contracts over a 20-year period report a Sharpe ratio of 1.74 and improvements in profitability and risk control. These findings are limited to the reported backtest; the document gives no further detail on costs, implementation, comparison methods, or out-of-sample validation. It also notes that the conventional network-building process can depend on costly databases and specialist knowledge, though it does not specify how those access barriers affect the reported results.

Key ideas

• Network momentum uses connections among assets to inform return predictions.
• L2GMOM jointly learns a financial network and the trading signals for a momentum strategy.
• Its neural network architecture is based on algorithm unrolling and is presented as interpretable.
• The framework can be trained with portfolio objectives such as the negative Sharpe ratio.
• The reported backtest covers 64 continuous futures contracts over 20 years and gives a Sharpe ratio of 1.74.

Tags

# Learning to Learn Financial Networks for Optimising Momentum Strategies


# Learning to Learn Financial Networks for Optimising Momentum Strategies









Network momentum provides a novel type of risk premium, which exploits the interconnections among assets in a financial network to predict future returns. However, the current process of constructing financial networks relies heavily on expensive databases and financial expertise, limiting accessibility for small-sized and academic institutions. Furthermore, the traditional approach treats network construction and portfolio optimisation as separate tasks, potentially hindering optimal portfolio performance. To address these challenges, we propose L2GMOM, an end-to-end machine learning framework that simultaneously learns financial networks and optimises trading signals for network momentum strategies. The model of L2GMOM is a neural network with a highly interpretable forward propagation architecture, which is derived from algorithm unrolling. The L2GMOM is flexible and can be trained with diverse loss functions for portfolio performance, e.g. the negative Sharpe ratio. Backtesting on 64 continuous future contracts demonstrates a significant improvement in portfolio profitability and risk control, with a Sharpe ratio of 1.74 across a 20-year period.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.