Modular meta-learning
Ferran Alet, Tomás Lozano‐Pérez, Leslie Pack Kaelbling · arXiv (Cornell University) · 2018
Many prediction problems, such as those that arise in the context of robotics, have a simplifying underlying structure that, if known, could accelerate learning. In this paper, we present a strategy for learning a set of neural network modules that can be combined in different ways. We train different modular structures on a set of related tasks and generalize to new tasks by composing the learned modules in new ways. By reusing modules to generalize we achieve combinatorial generalization, akin to the "infinite use of finite means" displayed in language. Finally, we show this improves performance in two robotics-related problems.
Cited by 29 works. doi:10.48550/arxiv.1806.10166 Open-access PDF
Authors
References
- Deep Residual Learning for Image Recognition
- Optimization by Simulated Annealing
- Adam: A Method for Stochastic Optimization
- Automatic differentiation in PyTorch
- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
- Learning to Learn
- Relational inductive biases, deep learning, and graph networks
- Neural Module Networks
- How to Grow a Mind: Statistics, Structure, and Abstraction
- PathNet: Evolution Channels Gradient Descent in Super Neural Networks