Learning Embedding Adaptation for Few-Shot Learning

By Han-Jia Ye, Hexiang Hu, De-chuan Zhan, Fei Sha. 2018

90 citations recorded by the source corpus.

Subfields: 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 Representation & Self-Supervised Learning

Citations in the corpus

Rethinking Few-Shot Image Classification: a Good Embedding Is All You Need? (2020, background): “rning measures a model’s ability to quickly adapt to new environments and tasks. This is a challenging problem because only limited data is available to adapt the model. Recently, signiﬁcant advances [56,55,53,12, 46,48,57,34,44,62,26,28] have been made to tackle this problem using the ideas of meta”
Diversity With Cooperation: Ensemble Methods for Few-Shot Classification (2019, neutral): “Speciﬁcally, better results were obtained by training the network on the classical classiﬁ-cation task using the meta-training data in a ﬁrst step, and then only ﬁne-tuning with meta-learning in a second step [22, 28, 34].”
Learning Adaptive Classifiers Synthesis for Generalized Few-Shot Learning (2019, methodology): “not inﬂuence the fewshot classiﬁcation performance a lot, but it is essential to make the GFSL classiﬁer well-calibrated. 4.3 Feature Network Speciﬁcation Following the setting of most recent methods [37, 44, 60], we use ResNet-12 [4, 18] to implement the embedding backbone ˚. Different from the sta”
Meta-Transfer Learning Through Hard Tasks (2019, methodology): “In this journal version, we implement two deeper architectures – ResNet-18 and ResNet-25, which have been adopted in newly published related works [42], [57], [72], and we achieve the top”
An Ensemble of Epoch-Wise Empirical Bayes for Few-Shot Learning (2019, methodology): “rchitecture. For MAML and MAML++, we use exactly the same network architecture as the original papers: a 4-layer convolution network (4CONV). For MTL, we have two options, ResNet-18 [24] and ResNet25 [32, 49]. Followed by convolution networks, we apply anaveragepoolinglayerandafully-connectedlayer. ”
Few-Shot Learning via Embedding Adaptation With Set-to-Set Functions (2018, methodology): “com/kjunelee/MetaOptNet 5We use the ResNet backbone with input image size 80 × 80 × 3 from [15] in the old version of our paper [30], whose source code of ResNet is publicly available on https://github.”
SimpleShot: Revisiting Nearest-Neighbor Classification for Few-Shot Learning (2019, background): “We compare our baselines to a range of state-of-the-art few-shot learners [1, 4, 5, 6, 8, 10, 15, 17, 21, 22, 23, 24, 26, 29, 25, 30, 31, 32, 34, 36].”
Leveraging the Feature Distribution in Transfer-based Few-Shot Learning (2020, methodology): “works [20, 18, 19, 13, 17, 9, 16, 10, 39], where we exploit unlabelled samples during the procedure as well as priors about their relative proportions.”
Learning Classifier Synthesis for Generalized Few-Shot Learning (2019, methodology): “ accuracy on MiniImageNet with the ResNet-12 backbone. Setups !1-Shot 5-Way 5-Shot 5-Way ProtoNet [ 29] 61.40 0.890.02 76.56 0.02 LEO [25] 61.76 0.08 77.59 0.12 OptNet [16] 62.64 0.61 78.63 0.46 FEAT [39] 62.96 0.720.02 78.49 0.02 Ours: C ASTLE63.51 0.020.02 79.52 0.02 Table 2: Few-shot classiﬁcatio”
Revisiting Mid-Level Patterns for Cross-Domain Few-Shot Recognition (2020, unknown): “For the CUB benchmark, we follow current works [31, 38, 45] to use the provided bounding boxes to crop the images.”
Heterogeneous Few-Shot Model Rectification With Semantic Mapping (2020, methodology): “Embedding [9], [11], model initialization [10], optimization conﬁgurations [51], classiﬁer mapping [52], and task adaptation strategy [19], [53] are validated to facilitate the cross-class transfer under the meta-learning paradigm.”
DeepEMD: Few-Shot Image Classification With Differentiable Earth Mover’s Distance and Structured Classifiers (2020, methodology): “Our method is more related to the metric-based methods [30,42,62,64,65,70], which aim to represent samples in an appropriate feature space where data from different categories can be distinguished with distance metrics.”
Memory-Based Neighbourhood Embedding for Visual Recognition (2019, methodology): “FEAT [46] transforms the embeddings from task-agnostic to task-specific by employing self-attention mechanism.”
Multimodal Model-Agnostic Meta-Learning via Task-Aware Modulation (2019, background): “Recent works along this line [29, 53, 19] share a conceptually similar idea with us and seek to perform task-specific adaptation with different type transformations.”
Few-shot learning with adaptively initialized task optimizer: a practical meta-learning approach (2019, methodology): “Different from that, in our experiments, we test the model over 10,000 sampled few-shot tasks (Rusu et al. 2018; Ye et al. 2018).”
Adaptive Subspaces for Few-Shot Learning (2020, background): “Other works use feature attention modules [28, 29] to modulate features for few-shot learning [30, 31].”
Modern Hopfield Networks and Attention for Immune Repertoire Classification (2020, background): “While this attention mechanism has originally been developed for sequence tasks (Vaswani et al., 2017), it can be readily transferred to sets (Lee et al., 2019; Ye et al., 2018).”
DeepEMD: Differentiable Earth Mover's Distance for Few-Shot Learning (2020, methodology): “in the state-of-the-art literature [3], we adopt a feature pre-”