Meta-Learning with Warped Gradient Descent

Abstract

Learning an efficient update rule from data that promotes rapid learning of new tasks from the same distribution remains an open problem in meta-learning. Typically, previous works have approached this issue either by attempting to train a neural network that directly produces updates or by attempting to learn better initialisations or scaling factors for a gradient-based update rule. Both of these approaches pose challenges. On one hand, directly producing an update forgoes a useful inductive bias and can easily lead to non-converging behaviour. On the other hand, approaches that try to control a gradient-based update rule typically resort to computing gradients through the learning process to obtain their meta-gradients, leading to methods that can not scale beyond few-shot task adaptation. In this work, we propose Warped Gradient Descent (WarpGrad), a method that intersects these approaches to mitigate their limitations. WarpGrad meta-learns an efficiently parameterised preconditioning matrix that facilitates gradient descent across the task distribution. Preconditioning arises by interleaving non-linear layers, referred to as warp-layers, between the layers of a task-learner. Warp-layers are meta-learned without backpropagating through the task training process in a manner similar to methods that learn to directly produce updates. WarpGrad is computationally efficient, easy to implement, and can scale to arbitrarily large meta-learning problems. We provide a geometrical interpretation of the approach and evaluate its effectiveness in a variety of settings, including few-shot, standard supervised, continual and reinforcement learning.

arXiv comment28 pages, 13 figures, 3 tables. Published as a conference paper at ICLR 2020

References in corpus (22)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• Prototypical Networks for Few-shot Learning
• A Simple Neural Attentive Meta-Learner
• Meta-SGD: Learning to Learn Quickly for Few-Shot Learning
• Learning multiple visual domains with residual adapters
• On First-Order Meta-Learning Algorithms
• Optimizing Neural Networks with Kronecker-factored Approximate Curvature
• Overcoming Catastrophic Forgetting by Incremental Moment Matching
• Efficient Off-Policy Meta-Reinforcement Learning via Probabilistic Context Variables
• Fast Context Adaptation via Meta-Learning
• Learning to Learn without Gradient Descent by Gradient Descent
• Using Fast Weights to Attend to the Recent Past
• Learning to Optimize
• Universal representations:The missing link between faces, text, planktons, and cat breeds
• Deep Meta-Learning: Learning to Learn in the Concept Space
• Meta-Learning Update Rules for Unsupervised Representation Learning
• Natural Neural Networks
• Transferring Knowledge across Learning Processes
• Backpropamine: training self-modifying neural networks with differentiable neuromodulated plasticity
• Guided Meta-Policy Search
• Uncertainty in Multitask Transfer Learning

Cited by in corpus (9)

• From Learning to Meta-Learning: Reduced Training Overhead and Complexity for Communication Systems
• FedBABU: Towards Enhanced Representation for Federated Image Classification
• Quantum Architecture Search with Meta-learning
• ReFine: Re-randomization before Fine-tuning for Cross-domain Few-shot Learning
• Scalable Bayesian Meta-Learning through Generalized Implicit Gradients
• Learning a Universal Template for Few-shot Dataset Generalization
• Meta Continual Learning via Dynamic Programming
• RotoGrad: Gradient Homogenization in Multitask Learning
• Uniform Sampling over Episode Difficulty