DreamCoder: Growing generalizable, interpretable knowledge with wake-sleep Bayesian program learning

TL;DR

DreamCoder is a system that learns to solve problems by creating and refining programming languages and neural networks, enabling it to understand and transfer complex concepts across diverse tasks.

Contribution

It introduces a wake-sleep Bayesian program learning framework that autonomously develops interpretable, compositional programming languages for problem-solving and concept learning.

Findings

Successfully solves classic inductive programming tasks

Discovers fundamental principles of physics and programming

Builds scalable, interpretable, and transferable symbolic representations

Abstract

Expert problem-solving is driven by powerful languages for thinking about problems and their solutions. Acquiring expertise means learning these languages -- systems of concepts, alongside the skills to use them. We present DreamCoder, a system that learns to solve problems by writing programs. It builds expertise by creating programming languages for expressing domain concepts, together with neural networks to guide the search for programs within these languages. A ``wake-sleep'' learning algorithm alternately extends the language with new symbolic abstractions and trains the neural network on imagined and replayed problems. DreamCoder solves both classic inductive programming tasks and creative tasks such as drawing pictures and building scenes. It rediscovers the basics of modern functional programming, vector algebra and classical physics, including Newton's and Coulomb's laws. Concepts are built compositionally from those learned earlier, yielding multi-layered symbolic representations that are interpretable and transferrable to new tasks, while still growing scalably and flexibly with experience.