Few-Shot Learning with Metric-Agnostic Conditional Embeddings

By Nathan Hilliard, Lawrence Phillips, Scott Howland, Artëm Yankov, Court D. Corley, Nathan Oken Hodas. 2018

165 citations recorded by the source corpus.

Learning high quality class representations from few examples is a key problem in metric-learning approaches to few-shot learning. To accomplish this, we introduce a novel architecture where class representations are conditioned for each few-shot trial based on a target image. We also deviate from traditional metric-learning approaches by training a network to perform comparisons between classes rather than relying on a static metric comparison. This allows the network to decide what aspects of each class are important for the comparison at hand. We find that this flexible architecture works well in practice, achieving state-of-the-art performance on the Caltech-UCSD birds fine-grained classification task.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Molecular Biology, Meta & Continual 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

1. Ming-Hsuan Yang
2. Dacheng Tao
3. Mingsheng Long
4. Yue Cao
5. Zheng Zhang
6. Mehrtash Harandi
7. Han Hu
8. Yun Raymond Fu

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Ming-Hsuan Yang
3. Dacheng Tao
4. Yi Yang
5. Xiaogang Wang
6. Chen Change Loy
7. Trevor Darrell
8. Andrea Vedaldi

In Molecular Biology

1. Hung-Yu Tseng
2. Ming-Hsuan Yang
3. Mingsheng Long
4. Yun Raymond Fu
5. Tong Zhang
6. Jia Li
7. Hsin-Ying Lee
8. Lie Wang

In Meta & Continual Learning

1. Christian Simon
2. Lars Petersson
3. Pengfei Fang
4. Mehrtash Harandi
5. Liang Song
6. Hung-Yu Tseng
7. Ming-Hsuan Yang
8. Yongqiang Qin

Citations in the corpus

• Further Advantages of Data Augmentation on Convolutional Neural Networks (2018, methodology): “The observation that data augmentation alone outperforms explicitly regularized models for few-shot learning was also made by Hilliard et al. in Hilliard et al. (2018).”
• A Two-Stage Approach to Few-Shot Learning for Image Recognition (2019, methodology): “Using the class splits in [60], we have 100, 50 and 50 categories used for training, validation and testing, respectively.”
• Negative Margin Matters: Understanding Margin in Few-shot Classification (2020, methodology): “Following the standard setting of [14], we split the classes in the dataset into 100 base classes, 50 validation classes, and 50 novel classes.”
• Fast and Generalized Adaptation for Few-Shot Learning (2019, methodology): “We followed the commonly used evaluation protocol proposed by [12].”
• Self-Supervised Learning for Few-Shot Image Classification (2019, unknown): “For evaluation, we follow the split in [27].”
• Reinforced Attention for Few-Shot Learning and Beyond (2021, methodology): “Following the data preparation from [15], we randomly divide it into 100 train classes, 50 validation classes and 50 test classes.”
• Cross-Domain Few-Shot Classification via Learned Feature-Wise Transformation (2020, methodology): “We follow the setting in Ravi & Larochelle (2017) and Hilliard et al. (2018) to process mini-ImageNet and CUB datasets.”
• Transductive Episodic-Wise Adaptive Metric for Few-Shot Learning (2019, methodology): “Follow the same partition as [9], we use 100 classes for training, and another two 50 classes as unseen classes for validation and evaluation.”
• Adversarial Feature Hallucination Networks for Few-Shot Learning (2020, methodology): “We use the split in [17] and 100, 50, 50 classes for training, validation, and testing, respectively.”