Learning to Compare: Relation Network for Few-Shot Learning

By Flood Sung, Yongxin Yang, Li Zhang, Tao A. Xiang, Philip H. S. Torr, Timothy M. Hospedales. 2017

4,711 citations recorded by the source corpus.

We present a conceptually simple, flexible, and general framework for few-shot learning, where a classifier must learn to recognise new classes given only few examples from each. Our method, called the Relation Network (RN), is trained end-to-end from scratch. During meta-learning, it learns to learn a deep distance metric to compare a small number of images within episodes, each of which is designed to simulate the few-shot setting. Once trained, a RN is able to classify images of new classes by computing relation scores between query images and the few examples of each new class without further updating the network. Besides providing improved performance on few-shot learning, our framework is easily extended to zero-shot learning. Extensive experiments on five benchmarks demonstrate that our simple approach provides a unified and effective approach for both of these two tasks.

Subfields: Artificial Intelligence, Cancer Research, Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Cancer Research

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

Citations in the corpus

Big Transfer (BiT): General Visual Representation Learning (2019, background): “Metalearning or metric-learning techniques have been proposed to learn with few or no labels [59,49,52].”
Rethinking Few-Shot Image Classification: a Good Embedding Is All You Need? (2020, methodology): “64 Relation Networks [48] 64-96-128-256 50.”
Repurposing Pretrained Models for Robust Out-of-domain Few-Shot Learning (2021, methodology): “First, metric-based approaches solve few-shot tasks by training a feature extractor that maximizes inter-class similarity and intra-class dissimilarity (Vinyals et al., 2016; Snell et al., 2017; Sung et al., 2018).”
CrossTransformers: spatially-aware few-shot transfer (2020, background): “Matching-based methods [28, 75, 78, 89] instead learn representations for similarity functions [11, 16, 17, 44, 82], in the hope that the similarities will generalize to new data.”
Meta-learning with differentiable closed-form solvers (2018, methodology): “Despite its simplicity, this general strategy is particularly effective and it is at the core of several state of the art few-shot classification algorithms [6, 10, 13].”
Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples (2019, methodology): “These were based on graph neural networks (Satorras & Estrach, 2018), relational networks (Sung et al., 2018) and attentional networks (Mishra et al., 2018).”
A Closer Look at Few-shot Classification (2019, methodology): “The problem of learning to generalize to unseen classes during training, known as few-shot clas-siﬁcation , has attracted considerable attention Vinyals et al. (2016); Snell et al. (2017); Finn et al. (2017); Ravi & Larochelle (2017); Sung et al. (2018); Garcia & Bruna (2018); Qi et al. (2018).”
Meta-Learning with Latent Embedding Optimization (2018, methodology): “Then, codes in H corresponding to different training examples are concatenated pair-wise (resulting in ( NK ) 2 pairs in the case of K -shot classiﬁcation) and processed by a relation network g φ r , in a similar fashion to Oreshkin et al. (2018) and Sung et al. (2017).”
Meta-Learning With Differentiable Convex Optimization (2019, neutral): “6 Relation Networks∗ [29] 64-96-128-256 55.”
Backdoor Attacks on Federated Meta-Learning (2020, background): “This may suggest a limitation of the classification mechanism of matching networks; other variants of matching networks, or models with non-parametric classification mechanisms such as Prototypical Networks [31] (computing distances to prototype representations of each class) and Relation Networks [”
Recasting Gradient-Based Meta-Learning as Hierarchical Bayes (2018, methodology): “Improved performance on miniImageNet has been reported by several works (Mishra et al., 2017; Munkhdalai & Yu, 2017; Sung et al., 2017) by making use of a model architecture with significantly more parameters than the methods in Table 1.”
Probabilistic Model-Agnostic Meta-Learning (2018, neutral): “36 % Relation Net [36] 50 .”
TADAM: Task dependent adaptive metric for improved few-shot learning (2018, background): “Since then, the problem of learning the most suitable metric for few-shot learning has been of interest to the field [33, 28, 29, 17, 16].”
Margin-Based Few-Shot Class-Incremental Learning with Class-Level Overfitting Mitigation (2022, methodology): “The latter achieves this goal by freezing parameters of the pre-trained model, and recognize the novel classes by the prototype-based [28] Nearest-Neighbor classification, which shares the same concept with the metric-based few-shot learning works [32, 28, 30].”
Forward Compatible Few-Shot Class-Incremental Learning (2022, methodology): “Metric-based algorithms utilize a pretrained backbone for feature extraction, and employ proper distance metrics between support and query instances [26, 37, 39, 45, 52, 53, 57].”
Generalized and Incremental Few-Shot Learning by Explicit Learning and Calibration without Forgetting (2021, background): “Generalized Few-Shot Learning (GFSL) Many approaches for few-shot learning (FSL) rely on a meta-learning paradigm to quickly adapt a method to new underrepresented samples [41, 47, 12, 43, 30, 43].”