Infinite Mixture Prototypes for Few-Shot Learning

By Kelsey R. Allen, Evan Shelhamer, Hanul Shin, Josh Tenenbaum. 2019

270 citations recorded by the source corpus.

We propose infinite mixture prototypes to adaptively represent both simple and complex data distributions for few-shot learning. Our infinite mixture prototypes represent each class by a set of clusters, unlike existing prototypical methods that represent each class by a single cluster. By inferring the number of clusters, infinite mixture prototypes interpolate between nearest neighbor and prototypical representations, which improves accuracy and robustness in the few-shot regime. We show the importance of adaptive capacity for capturing complex data distributions such as alphabets, with 25% absolute accuracy improvements over prototypical networks, while still maintaining or improving accuracy on the standard Omniglot and mini-ImageNet benchmarks. In clustering labeled and unlabeled data by the same clustering rule, infinite mixture prototypes achieves state-of-the-art semi-supervised accuracy. As a further capability, we show that infinite mixture prototypes can perform purely unsupervised clustering, unlike existing prototypical methods.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging, Probabilistic & Bayesian Methods, 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

In Computer Vision and Pattern Recognition

In Radiology, Nuclear Medicine and Imaging

In Probabilistic & Bayesian Methods

In Meta & Continual Learning

Citations in the corpus

Rethinking Few-Shot Image Classification: a Good Embedding Is All You Need? (2020, neutral): “73 IMP [2] 64-64-64-64 49.”
Novel Class Discovery without Forgetting (2022, background): “Few-shot learning methods [4,57,47] additionally require a few of the unlabeled instances to be labeled.”
Continuous Meta-Learning without Tasks (2019, methodology): “Replacing the Dirichlet prior with a Chinese restaurant process (as in e.g. Nagabandi et al. (2019b); Allen et al. (2019)) would enable a few-shot meta-classification model with a potentially expandable number of classes.”
Meta-Learning with Implicit Gradients (2019, background): “Meta-learning approaches can generally be categorized into metric-learning approaches that learn an embedding space where non-parametric nearest neighbors works well [29, 57, 54, 45, 3], black-box approaches that train a recurrent or recursive neural network to take datapoints as input”
Hyperspherical Prototype Networks (2019, unknown): “A few works have also investigated multiple prototypes per class [1, 29, 46].”
Adaptive Risk Minimization: A Meta-Learning Approach for Tackling Group Shift (2020, background): “Meta-learning [57, 6, 66, 28] has been most extensively studied in the context of few shot supervised learning methods [56, 69, 52, 18, 70, 62, 20, 1].”
Hypercorrelation Squeeze for Few-Shot Segmenation (2021, background): “To resolve the generalization problem, many recent approaches to image classification made various attempts in training deep networks with a few annotated examples [1, 18, 25, 31, 50, 53, 59, 65, 67, 73, 79, 84, 85].”
Wandering Within a World: Online Contextualized Few-Shot Learning (2020, methodology): “As described above, online IMP has the capability of learning from unlabeled examples, unlike online MatchingNet.”
Variable-Shot Adaptation for Online Meta-Learning (2020, background): “Recent works have proposed non-parametric metalearners that work effectively with variable numbers of shots (Allen et al., 2019b; Cao et al., 2020).”
Variational Few-Shot Learning (2019, methodology): “For instance, prototypebased methods [42, 15, 1] compute the mean of embedded support examples for each class, which is vulnerable to noise since data is severely limited.”
On Episodes, Prototypical Networks, and Few-shot Learning (2020, background): “[27, 18, 1, 50, 7]), and they are used in a number of applied machine learning works (e.”
Learning to Learn Kernels with Variational Random Features (2020, unknown): “…± 0.1 97.6 ± 0.3 99.4 ± 0.2 GNN (Garcia & Bruna, 2018) 99.2 99.7 97.4 99.0 VERSA (Gordon et al., 2019) 99.7 0.2 ± ± ± ± R2-D2 (Bertinetto et al., 2019) 98.6 99.7 94.7 98.9 IMP (Allen et al., 2019) 98 usually set to be 5 to 10 times the dimension of the input features (Yu et al., 2016; Rahimi &…”
PODNet: Pooled Outputs Distillation for Small-Tasks Incremental Learning (2020, methodology): “This version brings to mind SoftTriple [32] and Infinited Mixture Prototypes [2], used in the different context of few-shot learning.”