Modular Neural Architectures

Taken from “A Path Towards Autonomous Machine Intelligence”, by Yann LeCun

Motivation

Despite the success of Mixture-of-Expert Transformers, the modular design of the human brain enables far more flexibility.

I’m working on designing more flexiable Transformers with more flexible modular design.

Milestones

  • Learning Multi-Objective Curricula for Robotic Policy Learning, CoRL 2022. We propose a unified automatic curriculum learning framework to create multi-objective but coherent curricula that are generated by a set of parametric curriculum modules. Each curriculum module is instantiated as a neural network and is responsible for generating a particular curriculum.

Next

  • Make modular designs more scalable to handle large-scale benchmark tasks, e.g., MineDojo.
Jie Fu
Jie Fu
Researcher (Principal Investigator)

Focus on Deep Learning, AI for Science, Reinforcement Learning, Language Processing.