Boosting Supervision with Self-Supervision for Few-shot Learning

Abstract

We present a technique to improve the transferability of deep representations learned on small labeled datasets by introducing self-supervised tasks as auxiliary loss functions. While recent approaches for self-supervised learning have shown the benefits of training on large unlabeled datasets, we find improvements in generalization even on small datasets and when combined with strong supervision. Learning representations with self-supervised losses reduces the relative error rate of a state-of-the-art meta-learner by 5-25% on several few-shot learning benchmarks, as well as off-the-shelf deep networks on standard classification tasks when training from scratch. We find the benefits of self-supervision increase with the difficulty of the task. Our approach utilizes the images within the dataset to construct self-supervised losses and hence is an effective way of learning transferable representations without relying on any external training data.

References in corpus (9)

• Adam: A Method for Stochastic Optimization
• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• Prototypical Networks for Few-shot Learning
• Fine-Grained Visual Classification of Aircraft
• Matching Networks for One Shot Learning
• GradNorm: Gradient Normalization for Adaptive Loss Balancing in Deep Multitask Networks
• Few-Shot Learning with Graph Neural Networks
• Multi-Task Learning as Multi-Objective Optimization
• Revisiting Self-Supervised Visual Representation Learning

Cited by in corpus (4)

• Unsupervised Domain Adaptation through Self-Supervision
• Test-Time Training with Self-Supervision for Generalization under Distribution Shifts
• Few-shot Action Recognition with Permutation-invariant Attention
• Batch Group Normalization