Description

Kernels, Pricing, Attention
Nonparametric thinking has transformed econometrics, finance and artificial intelligence.
Kernels let the data speak, revealing economic structure beyond parametric restrictions.
Engel curves illustrate how flexible smoothing uncovers consumption behaviour.
Pricing kernels connect observed market prices to hidden investor preferences.
Option markets allow us to estimate stochastic discount factors nonparametrically.
The classical pricing-kernel puzzle challenges standard asset-pricing theory.
Modern methods—from kernels and splines to forests—provide adaptive estimators.
Transformers extend the same idea: learning through data-dependent weighting schemes.
Attention can be viewed as an adaptive kernel operating in learned representation space.
From Engel to Arrow–Debreu to GPT: kernels remain the common mathematical language.

In the beginning was the kernel.
It explained household demand through Engel curves,
revealed investor preferences through pricing kernels,
and today reappears as attention in large language models.
Nonparametric smoothing became asset pricing,
asset pricing became representation learning,
and representation learning became artificial intelligence.
The mathematics changed its clothes,
but the central idea remained the same:
learn by weighting information locally and adaptively.