Past Recording
Discovering Symbolic Inductive Biases
Wednesday Aug 19 2020 16:00 GMT
Please to join the live chat.
Discovering Symbolic Inductive Biases
Why This Is Interesting

The authors present an approach offer alternative directions for interpreting neural networks and discovering novel physical principles from the representations they learn. This is accomplished by applying symbolic regression to components of the learned model to extract explicit physical relations. This is also applied to a non-trivial cosmology example—a detailed dark matter simulation—and discover a new analytic formula which can predict the concentration of dark matter from the mass distribution of nearby cosmic structures.

Discussion Points
  • What is symbolic regression?
  • What does it mean for a DL model to have a separable internal structure?
  • How do you deal with the curse of dimensionality?
Time of Recording: Wednesday Aug 19 2020 16:00 GMT