Deep Tangency Portfolios

We propose a parametric approach for directly estimating the tangency portfolio weights by combining fundamental finance theory and deep learning techniques. The deep tangency portfolio is a combination of the market factor and a deep long-short factor constructed using a large number of characteristics. We apply our approach to the corporate bond market. Albeit acting as a market-hedged portfolio, the deep factor achieves a sizable risk premium with an out-of-sample annualized Sharpe ratio of 2.08. The deep tangency portfolio outperforms those constructed from commonly used observable or latent factors with an out-of-sample annualized Sharpe ratio of 2.90.