Discovering Symbolic Models from Deep Learning with Inductive Biases

Discovering Symbolic Models from Deep Learning with Inductive Biases

Discovering Symbolic Models from Deep Learning with Inductive Biases

“To automate science we need to automate knowledge discovery. However, when does a machine learning model become knowledge? Why are Maxwell’s equations considered a fact of science, but a deep learning model just an interpolation of data? For one, deep learning doesn’t generalize near as well as symbolic physics models…”

Source: astroautomata.com/paper/symbolic-neural-nets/

Subscribe to our Digest