Rapid Adaptation with Conditionally Shifted Neurons

By Tsendsuren Munkhdalai, Xingdi Yuan, Soroush Mehri, Adam Trischler. 2017

325 citations recorded by the source corpus.

How researchers see this paper

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

In Meta & Continual Learning

Citations in the corpus

Meta-Learning With Differentiable Convex Optimization (2019, neutral): “Model-based methods [18, 19] learn a parameterized predictor to estimate model parameters, e.”
Uncertainty in Multitask Transfer Learning (2018, background): “…[31] 60.0 % Meta-Learner LSTM [28] 60.6 % MAML [11] 63.2% Prototypical Networks [29] 68.2 % SNAIL [24] 68.9 % Discriminative k-shot [3] 73.9 % adaResNet [25] 71.9 % Deep Prior (Ours) 62.7 % Deep Prior + Proto Net (Ours) 74.5 % We ﬁrst experimented with the vanilla version of deep prior (2).”
Meta-Transfer Learning for Few-Shot Learning (2018, methodology): “miniImageNet is widely used in related works [9, 38, 13, 11, 31].”
Learning to Self-Train for Semi-Supervised Few-Shot Classification (2019, background): “They are based on either data augmentation [15, 29] or gradient descent [3, 23, 7, 5, 42, 19, 27, 32, 10].”
Meta-Curvature (2019, methodology): “, 2018) with larger ResNet models (Qiao et al., 2018; Rusu et al., 2019; Munkhdalai et al., 2018; Mishra et al., 2018).”
Learning to Adapt in Dynamic, Real-World Environments through Meta-Reinforcement Learning (2018, background): “…approach involves simply unrolling a recurrent neural network (RNN) that ingests the data (Santoro et al., 2016; Munkhdalai and Yu, 2017; Munkhdalai et al., 2017; Mishra et al., 2017) and learns internal representations of the algorithms themselves, one instantiation of our approach…”
Generalizing from a Few Examples (2019, neutral): “f1(x(i)), f2(x(i))] y(i)and age dot product MN-Net [19] f (x(i)) y(i) dot product life-long memory [52] f (x(i)) y(i)and age cosine similarity MetaNet [80] f (x(i)) fast weight cosine similarity CSNs [81] f (x(i)) fast weight cosine similarity APL [89] f (x(i)) y(i) squared ℓ2 distance Table 5 intro”
Meta-Transfer Learning Through Hard Tasks (2019, methodology): “respectively for sampling tasks for meta-training, meta-validation and meta-test, following related works [7], [20], [21], [45], [70].”
An Ensemble of Epoch-Wise Empirical Bayes for Few-Shot Learning (2019, methodology): “.1. Datasets and implementation details We conduct few-shot learning experiments on two benchmarks: miniImageNet [59] and Fewshot-CIFAR100 (FC100) [40]. The former one is widely used in related works [11, 43, 17, 13, 38, 54], and the later one is more challenging due to lower image resolution and ha”
Adaptive Subspaces for Few-Shot Learning (2020, unknown): “AdaResNet [50] ResNet-12 56.”
On Episodes, Prototypical Networks, and Few-shot Learning (2020, methodology): “Pioneered by the seminal work of Utgoff (1986), Schmidhuber (1987, 1992), Bengio et al.”
A New Meta-Baseline for Few-Shot Learning (2020, background): “(Munkhdalai et al., 2017) modifies the activation values of a network by shifting them according to the task-specific information.”
Learning to Adapt: Meta-Learning for Model-Based Control (2018, unknown): “Another popular meta-learning approach involves simply unrolling a recurrent neural network (RNN) that ingests the data (Santoro et al., 2016; Munkhdalai & Yu, 2017; Munkhdalai et al., 2017; Mishra et al., 2017) and learns internal representations of the algorithms themselves.”
A Two-Stage Approach to Few-Shot Learning for Image Recognition (2019, background): “Similarly, Munkhdalai et al. [37] proposed a meta-learning scheme that shifts the neuron activations depending on task-speciﬁc parameters.”
Few-Shot Object Detection With Attention-RPN and Multi-Relation Detector (2019, background): “This research direction is named meta-learning in general [2, 5, 37, 38, 39].”
Cross Attention Network for Few-shot Classification (2019, methodology): “Following [26, 17, 36, 21], we use ResNet-12 network as our embedding module.”
One-Shot Instance Segmentation (2018, methodology): “pproach has seen a number of improvements in recent years [36, 84, 79, 85, 86]. Other approaches are based on generative models [38, 76], ideas from information retrieval [81] or employ meta learning [24, 40, 52, 51, 53, 54, 74, 80, 69]. Few-shot segmentation. Closely related to our work is one-shot”
Learning to Learn Kernels with Variational Random Features (2020, unknown): “Note that, in ± (Gidaris & Komodakis, 2018) 56.20 0.86 ± 73.00 0.64 ± (Bauer et al., 2017) 56.30 0.40 ± 73.90 0.30 ± ± (Munkhdalai et al., 2017) 57.10 0.70 70.04 0.63 ± (Qiao et al., 2018) 59.60 0.41 ± 73.54 0.19 ± ± LEO (Rusu et al., 2019) 61.76 0.08 77.59 0.12 ± SNAIL (Mishra et al., 2018) 55.71…”