Fast and Flexible Multi-Task Classification Using Conditional Neural Adaptive Processes

By James Requeima, Jonathan Gordon, John F. Bronskill, Sebastian Nowozin, Richard E. Turner. 2019

265 citations recorded by the source corpus.

The goal of this paper is to design image classification systems that, after an initial multi-task training phase, can automatically adapt to new tasks encountered at test time. We introduce a conditional neural process based approach to the multi-task classification setting for this purpose, and establish connections to the meta-learning and few-shot learning literature. The resulting approach, called CNAPs, comprises a classifier whose parameters are modulated by an adaptation network that takes the current task's dataset as input. We demonstrate that CNAPs achieves state-of-the-art results on the challenging Meta-Dataset benchmark indicating high-quality transfer-learning. We show that the approach is robust, avoiding both over-fitting in low-shot regimes and under-fitting in high-shot regimes. Timing experiments reveal that CNAPs is computationally efficient at test-time as it does not involve gradient based adaptation. Finally, we show that trained models are immediately deployable to continual learning and active learning where they can outperform existing approaches that do not leverage transfer learning.

Subfields: Artificial Intelligence, 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 Meta & Continual Learning

Citations in the corpus

Flamingo: a Visual Language Model for Few-Shot Learning (2022, methodology): “In this paper, we take inspiration from the in-context [11] few-shot learning technique instead of more involved few-shot learning approaches based on metric learning [24, 103, 112, 117] or meta-learning [6, 7, 27, 31, 91, 155].”
Few-Shot Parameter-Efficient Fine-Tuning is Better and Cheaper than In-Context Learning (2022, methodology): “[79] use conditional neural adaptive processes and Li et al.”
ORBIT: A Real-World Few-Shot Dataset for Teachable Object Recognition (2021, methodology): “This, along with the total number of parameters (which are similar across models), suggests that ProtoNets and CNAPs would be better suited to deployment on a mobile device.”
CrossTransformers: spatially-aware few-shot transfer (2020, neutral): “55 CNAPS [65] 50.”
Learn Continually, Generalize Rapidly: Lifelong Knowledge Accumulation for Few-shot Learning (2021, methodology): “Generally, existing CL methods encompass memory and generative replaybased approaches (Robins, 1995; Lopez-Paz and Ranzato, 2017; Shin et al., 2017), regularization based approaches (Kirkpatrick et al., 2017; Nguyen et al., 2018) and model expansion based approaches (Shin et al., 2017).”
Generalized Variational Continual Learning (2020, methodology): “, 2017), multi-task meta-learning (Requeima et al., 2019), and few-shot learning (Perez et al.”
Meta-Learning in Neural Networks: A Survey (2020, methodology): “Embedding the support set is often achieved by recurrent networks [51], [112], [113] convolution [39], or set embeddings [45], [108].”
LEEP: A New Measure to Evaluate Transferability of Learned Representations (2020, methodology): “The last row of Table 1 shows the correlation coefﬁcient between LEEP scores and test accuracies of CNAPs.”
Adaptive Risk Minimization: A Meta-Learning Approach for Tackling Group Shift (2020, methodology): “We demonstrate in the next section how paradigms such as contextual meta-learning [16, 52] are readily extended using the ARM framework.”
How to Train Your MAML to Excel in Few-Shot Classification (2021, background): “Another direction is to enable task-specific initialization [75, 99, 109, 111], which often needs additional sub-networks.”
TaskNorm: Rethinking Batch Normalization for Meta-Learning (2020, methodology): “Next, we evaluate NLs on a demanding few-shot classi-ﬁcation challenge called Meta-Dataset, composed of thirteen (eight train, ﬁve test) image classiﬁcation datasets (Tri-antaﬁllou et al., 2020).”
Memory Efficient Meta-Learning with Large Images (2021, methodology): “CNAPS [4] and VERSA [17]), the aggregation of support set points is built in directly via a deep set encoder eφ1(·).”
Improved Few-Shot Visual Classification (2019, unknown): “For an in-depth explanation, refer to the original paper [30].”
Bayesian Few-Shot Classification with One-vs-Each Pólya-Gamma Augmented Gaussian Processes (2020, background): “CNAPs (Requeima et al., 2019) modulate taskspecific Feature-wise Linear Modulation (FiLM) (Perez et al.”
Learning a Universal Template for Few-shot Dataset Generalization (2021, methodology): “We compare against CNAPs (Requeima et al., 2019), TaskNorm (Bronskill et al., 2020) 3 , SimpleCNAPS (Bateni et al., 2020), SUR(-pf) (Dvornik et al., 2020) and URT(-pf) (Liu et al., 2021).”
A Broader Study of Cross-Domain Few-Shot Learning (2019, background): “Recently, [47] proposes a successful meta-learning approach based on conditional neural process on the MetaDataset benchmark.”
Orthogonal Projection Loss (2021, unknown): “Similar to [42], we expand Meta-Dataset by adding three additional datasets, MNIST, CIFAR10, and CIFAR100.”
FiT: Parameter Efficient Few-shot Transfer Learning for Personalized and Federated Image Classification (2022, methodology): “CNAPs [9] also uses a frozen, pretrained backbone with FiLM layers added.”
A Universal Representation Transformer Layer for Few-Shot Image Classification (2020, background): “strategies [17] and more flexible meta-learning algorithms [16].”