Introducing Deep Learning in Asset Pricing for after-work quants

Takeaway. Asset pricing wants a stochastic discount factor (SDF) that prices the cross-section; Chen–Pelger–Zhu let a deep net (with an adversarial / no-arbitrage flavour) learn that kernel from characteristics—research architecture, not a fund.

Deep Learning in Asset Pricing (Luyang Chen, Markus Pelger, Jason Zhu, 2019) is the after-work classic for “skip the linear factor zoo, estimate an SDF.”

Fit for learners: Classic end-to-end GAN/SDF deep pricing — great after-work Python project to map characteristics to returns without a linear factor zoo.

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

The problem

Linear factor models struggle when true risk prices are nonlinear functions of many characteristics. Hand-picking factors invites the zoo and omitted-variable debates.

The idea / core intuition

• Represent the SDF / pricing kernel as a neural function of information (characteristics, macro states).
• Train so pricing errors (moment conditions) shrink—often with a generator/adversary flavour rather than plain MSE on returns.
• The output is a model of discounting, not a stock tip list.
• Compare implied factors / Sharpe of the SDF against linear benchmarks in the paper’s tables—with costs and OOS splits in mind.

Research pattern—not a recommendation to long the SDF portfolio live.

Why after-work Python learners care

This is the conceptual leap from “predict returns with XGBoost” to “price assets with a kernel.” You still need Python loops, tensors, and honest date cuts.

Tiny evening exercise: write a 3-asset toy excess-return vector and a candidate SDF m = 1 - b*r_mkt; print mean m * r for each asset; write one sentence on what “pricing error ≈ 0” would mean—learning only.

One-evening reading checklist

1. Define SDF in one sentence (prices payoffs by discounting).
2. State what the network is approximating.
3. Note why an adversarial / moment objective differs from return MSE.
4. List one linear factor baseline you would still plot.

Download PDF

Download PDF

Landing: arXiv abs

Important disclaimer

Not investment advice. This Hub note is educational only. The paper is academic asset-pricing research—not a recommendation to trade deep SDF portfolios with real capital.

What this is not

• Not CEF / subsidy framing
• Not Quant or Macro course CTA
• Not a live hedge-fund pitch

Next step

Single course CTA: SoloQuant Academy’s Introduction to Python Programming—functions and arrays first; SDF tensors wait until the syntax is boring.