Retro-engineering the US tariffs formula with deep reinforcement learning 👀
Автор: Wassim Tenachi
Загружено: 2025-04-08
Просмотров: 1713
Описание:
Follow to learn more about automated equation discovery ! ⚙️
Can AI guess the trade tariff formula just by looking at the data? 👀
Here’s Φ-SO - the Physical Symbolic Optimization algorithm - rediscovering the tariff law implemented in the US — directly from data.
No prior knowledge. Just learning patterns.
Φ-SO uses deep reinforcement learning to discover interpretable symbolic equations from data by trial-and-error — equations that make sense to humans.
In this case, it reverse-engineers the tariff formula from trade data alone.
More than curve-fitting, Φ-SO searches in the space of mathematical expressions themselves - guided by a neural policy and physical priors (e.g. units, symmetries).
It’s already been used in physics, cosmo, biology, aero/hydro sims — and now, economics!
For more details, have a look at the paper or the code! :)
code: github.com/WassimTenachi/PhySO
paper: arxiv.org/abs/2303.03192
tariffs demo: /demos/sr/demo_usa_tariffs
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
![Python Symbolic Regression (PySR) [Physics Informed Machine Learning]](https://imager.clipsaver.ru/df43V4OjMVs/max.jpg)
