[11] Jacob Andreas - Learning from Language
The Thesis Review
00:00
Deep Networks as Functional Programs
If you have primitives and rules for combining them so if you have a parse then you can get a network to kind of learn to use that structure in a sense yeah. The other thing that I wanted to say on the functional programs point is that one of the things that like sat on my to-do list literally for years isCan we use these models as models of programming language data and build basically abstract interpreters? It's been fun working with Armando Solars, Lazama students on actually applying these kinds of module net models to functional programs.
Transcript
Play full episode
