Introducing Deep Learning for Predicting Asset Returns for after-work quants

Takeaway. Expected-return prediction is a high-dimensional characteristics problem; deep nets can nest linear factor models as a special case—use it as a comparison template against OLS, not as a promised alpha engine.

Deep Learning for Predicting Asset Returns (Guanhao Feng, Jingyu He, Nicholas Polson, 2018) is an after-work bridge from linear cross-section regressions to flexible nets.

Fit for learners: Open-PDF classic/working paper on ml-finance — useful for after-work Python quant learners building intuition before coding.

This Hub note is a plain-language intro. Not investment advice. Keep Harvey–Liu–Zhu nearby.

The problem

Linear “characteristic → return” regressions miss interactions and thresholds. Adding more factors by hand is the zoo. Learners need a story for why a net might help—and why it overfits faster than OLS.

The idea / core intuition

• Map firm characteristics through hidden layers to predicted returns (or related targets).
• View depth as a way to compose nonlinear factors, with linear models nested inside.
• Regularisation, architecture, and time-respecting splits matter more than layer count slogans.
• Report against linear / shallow baselines; a fancier net that loses to OLS is a lesson, not a failure of the reader.

Research pattern—not a stock-picking product.

Why after-work Python learners care

If you can run LinearRegression on a characteristics matrix, this paper is permission to ask what a two-layer net changes—and what it doesn’t.

Tiny evening exercise: make a 100×5 random characteristic matrix and a random return vector; print OLS R² vs a dummy “always predict mean”; write why in-sample R² is not a trading edge—learning only.

One-evening reading checklist

1. State the input (characteristics) and output (return prediction).
2. Note how a deep net can nest a linear model.
3. List two evaluation pitfalls (leakage, multiple testing).
4. Name one baseline you would keep on the same chart as the net.

Download PDF

Download PDF

Landing: arXiv abs

Important disclaimer

Not investment advice. This Hub note is educational only. The paper is academic ML research—not a recommendation to trade predicted-return nets with real capital.

What this is not

• Not CEF / subsidy framing
• Not Quant or Macro course CTA
• Not an “AI stock picker” product pitch

Next step

Single course CTA: SoloQuant Academy’s Introduction to Python Programming—matrices start as lists of lists; then sklearn/PyTorch weekends make more sense.