A Closer Look at Few-shot Classification

By Wei-Yu Chen, Yen-Cheng Liu, Zsolt Kira, Yu-Chiang Frank Wang, Jia-Bin Huang. International Conference on Learning Representations, 2019

2,007 citations recorded by the source corpus.

Few-shot classification aims to learn a classifier to recognize unseen classes during training with limited labeled examples. While significant progress has been made, the growing complexity of network designs, meta-learning algorithms, and differences in implementation details make a fair comparison difficult. In this paper, we present 1) a consistent comparative analysis of several representative few-shot classification algorithms, with results showing that deeper backbones significantly reduce the performance differences among methods on datasets with limited domain differences, 2) a modified baseline method that surprisingly achieves competitive performance when compared with the state-of-the-art on both the \miniI and the CUB datasets, and 3) a new experimental setting for evaluating the cross-domain generalization ability for few-shot classification algorithms. Our results reveal that reducing intra-class variation is an important factor when the feature backbone is shallow, but not as critical when using deeper backbones. In a realistic cross-domain evaluation setting, we show that a baseline method with a standard fine-tuning practice compares favorably against other state-of-the-art few-shot learning algorithms.

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

View Paper: https://arxiv.org/abs/1904.04232

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 Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

Citations in the corpus

Do Better ImageNet Models Transfer Better? (2018, background): “Common few-shot learning methods do not seem to outperform classifiers trained on fixed features when domain shift is present [8], but it may be possible to obtain better results with specialized methods [18] or by combining few-shot learning methods with fine-tuning [60].”
Big Transfer (BiT): General Visual Representation Learning (2019, background): “However, recent work has shown that a simple linear classifier on top of pre-trained representations or fine-tuning can attain similar or better performance [7,37].”
Rethinking Few-Shot Image Classification: a Good Embedding Is All You Need? (2020, methodology): “Besides, [5] shows an improved fine-tuning model performs slightly worse than meta-learning algorithms.”
Shaping Visual Representations with Language for Few-Shot Classification (2019, neutral): “We evaluate on the 5 -way, 1 -shot setting, and as f θ use the 4-layer convolutional backbone used in much of the few-shot literature [4].”
Meta-Dataset: A Dataset of Datasets for Learning to Learn from Few Examples (2019, methodology): “Also, since Chen et al. (2019) added weight normalization (Salimans & Kingma, 2016) to their implementation of the cosine classifier layer, we also implemented this and created a hyperparameter choice for whether or not it is enabled.”
Backdoor Attacks on Federated Meta-Learning (2020, methodology): “We adopt a variant of matching networks from [32] where (1) the output components of the embedding model fθ are multiplied by gate variables 0 6 αl,j 6 1, and (2) cosine distances to reference examples of each class are multiplied by a scaling factor βj .”
Adaptive Gradient-Based Meta-Learning Methods (2019, neutral): “[17] Wei-Yu Chen, Yen-Cheng Liu, Zsolt Kira, Yu-Chiang Frank Wang, and Jia-Bin Huang.”
Diversity With Cooperation: Ensemble Methods for Few-Shot Classification (2019, methodology): “There is indeed strong evidence [5, 11, 23] that training CNNs from scratch using meta-learning performs substantially worse than if CNN features are trained in a standard fashion—that is, by minimizing a classical loss function relying on corpus annotations; on the other hand, learn-”
Few-Shot Continual Active Learning by a Robot (2022, methodology): “We compare our approach against 6 continual learning approaches (LWF [27], EWC [4], CWR [22], iCaRL [3], NCM [11], CBCL [10]), finetuning (FT) and a few-shot batch learning baseline (FLB) [28].”
Meta-Learning with Less Forgetting on Large-Scale Non-Stationary Task Distributions (2022, neutral): “We follow the same split of [17].”
Incremental Few-Shot Semantic Segmentation via Embedding Adaptive-Update and Hyper-class Representation (2022, methodology): “The former methods [6, 10, 20, 36] generally define few-shot learning as "learning to finetune", which aims to learn proper model initialization or predict network parameters.”
Few-Shot Class-Incremental Learning by Sampling Multi-Phase Tasks (2022, background): “Few-Shot Learning aims to extract the inductive bias from base classes and generalize it to unseen classes [44], [45], [46], [47], [48].”
Doodle It Yourself: Class Incremental Learning by Drawing a Few Sketches (2022, background): “Thus, off-the-shelf models naively pre-trained on photos commonly fail to generalise well on sketches [8].”
Forward Compatible Few-Shot Class-Incremental Learning (2022, background): “Few-Shot Learning (FSL): aims to fit unseen classes with insufficient training instances [9,47].”
Learngene: From Open-World to Your Learning Task (2021, methodology): “…of eight representative methods ((Finn, Abbeel, and Levine 2017; Nichol, Achiam, and Schulman 2018; Snell, Swersky, and Zemel 2017; Triantaﬁllou et al. 2020; Vinyals et al. 2016; Zhang et al. 2020), two ﬁne-tuning methods of (Chen et al. 2019)) to match ours on the data of novel classes.”
Few-Shot Incremental Learning with Continually Evolved Classifiers (2021, background): “Few-shot learning aims to learn a model that can classify unseen images when only training from scarce labeled training examples [5,48].”
Memory-Efficient Incremental Learning Through Feature Adaptation (2020, methodology): “Some recent methods [9,37] train neural networks with a small number of labeled examples.”