Portfolio construction through deep reinforcement learning and interpretable AI

Researchers from Cornell, Tsinghua and Beihang universities directly optimize the objectives of portfolio management via reinforcement learning — an alternative to conventional supervised learning-based paradigms that entail first-step estimations of return distributions, pricing kernels, or risk premia.

Building upon breakthroughs in artificial intelligence (AI), they develop multi-sequence neural network models tailored to distinguishing features of economic and financial data, while allowing training without labels and potential market interactions. The resulting AlphaPortfolio yields stellar out-of-sample performances (e.g., Sharpe ratio above two and over 13% risk-adjusted alpha with monthly re-balancing) that are robust under various economic restrictions and market conditions (e.g., exclusion of small stocks and short selling).

They project AlphaPortfolio onto simpler modeling spaces (e.g., using polynomial-feature-sensitivity) to uncover key drivers of investment performance, including their rotation and nonlinearity. More generally, they highlight the utility of deep reinforcement learning in finance and invent “economic distillation” tools for interpreting AI and big data models.


Notes: SREM refers to sequence representation extraction models (SREM), such as Transformer Encoder (TE) and Long Short-Term Memory (LSTM); CAAN refers to cross asset attention networks. Source: AlphaPortfolio: Direct Construction Through Deep Reinforcement Learning and Interpretable AI


Access the full paper

Related Posts

Previous Post
Bank of England’s Andrew Hauser on principles for central bank intervention
Next Post
NIST: machine learning for access control policy verification

Related Posts

No results found.

Fill out this field
Fill out this field
Please enter a valid email address.


Reset password

Create an account