BriefGPT.xyz
Jun, 2021
可扩展神经符号回归
Neural Symbolic Regression that Scales
HTML
PDF
Luca Biggio, Tommaso Bendinelli, Alexander Neitz, Aurelien Lucchi, Giambattista Parascandolo
TL;DR
本研究介绍了第一种使用大规模预训练的符号回归方法，通过生成一组不受限方程式并使用Transformer预测输入输出对应的符号方程，提高了方程式发现的效率和准确性。
Abstract
Symbolic equations are at the core of scientific discovery. The task of discovering the underlying equation from a set of input-output pairs is called
symbolic regression
. Traditionally,
symbolic regression
metho
→