Probabilistic Model-Agnostic Meta-Learning

By Chelsea Finn, Kelvin Xu, Sergey Levine. 2018

740 citations recorded by the source corpus.

Meta-learning for few-shot learning entails acquiring a prior over previous tasks and experiences, such that new tasks be learned from small amounts of data. However, a critical challenge in few-shot learning is task ambiguity: even when a powerful prior can be meta-learned from a large number of prior tasks, a small dataset for a new task can simply be too ambiguous to acquire a single model (e.g., a classifier) for that task that is accurate. In this paper, we propose a probabilistic meta-learning algorithm that can sample models for a new task from a model distribution. Our approach extends model-agnostic meta-learning, which adapts to new tasks via gradient descent, to incorporate a parameter distribution that is trained via a variational lower bound. At meta-test time, our algorithm adapts via a simple procedure that injects noise into gradient descent, and at meta-training time, the model is trained such that this stochastic adaptation procedure produces samples from the approximate model posterior. Our experimental results show that our method can sample plausible classifiers and regressors in ambiguous few-shot learning problems. We also show how reasoning about ambiguity can also be used for downstream active learning problems.

Subfields: Artificial Intelligence, 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 Probabilistic & Bayesian Methods

In Meta & Continual Learning

Citations in the corpus

Meta-Learning Probabilistic Inference for Prediction (2018, background): “In Section 4, we show that ML-PIP re-frames and extends existing point-estimate probabilistic interpretations of meta-learning (Grant et al., 2018; Finn et al., 2018) to cover a broader class of methods, including gradient based meta-learning (Finn et al.”
Learning to Learn with Generative Models of Neural Network Checkpoints (2022, background): “Subsequent work has built simple probabilistic models over learned MAML initializations (Finn et al., 2018).”
Meta-Learning with Latent Embedding Optimization (2018, methodology): “Concurrently with our work, Kim et al. (2018) and Finn et al. (2018) propose probabilistic extensions to MAML that are trained using a variational approximation, using simple posteriors.”
A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms (2019, neutral): “, 2017), and its related algorithms (Grant et al., 2018; Kim et al., 2018; Finn et al., 2018).”
Meta-Learning with Less Forgetting on Large-Scale Non-Stationary Task Distributions (2022, neutral): “Various approaches have been proposed [56,54,21,22,24,50,5,40,4,39,34,66,69,65].”
Few-Shot Class-Incremental Learning by Sampling Multi-Phase Tasks (2022, background): “MAML is widely applied into vast learning scenarios, e.g., robotics con-trol [49], neural translation [50] and uncertainty estimation [51], and many improved algorithms [52], [53], [54], [55], [56] have been made to further boost its performance.”
Embracing Change: Continual Learning in Deep Neural Networks. (2020, background): “In [86,87] both loops use gradient descent, while inner-loops can also be nearest-neighbor matching [88,89], recurrent neural representations [8,90,91], probabilistic inference [82,92], or mixed approaches [93–96].”
Neural Processes (2018, background): “…but there is a large literature of similar models for classiﬁcation (Koch et al., 2015; Santoro et al., 2016), reinforcement learning (Wang et al., 2016), parameter update (Finn et al., 2017; 2018), natural language processing (Bowman et al., 2015) and program induction (Devlin et al., 2017).”
Meta-Transfer Learning for Few-Shot Learning (2018, background): “Its task-agnostic property makes it possible to generalize to few-shot supervised learning as well as unsupervised reinforcement learning [13, 10].”
Rapid Learning or Feature Reuse? Towards Understanding the Effectiveness of MAML (2019, methodology): “This algorithm, with deep neural networks as the underlying model, has been highly influential, with significant follow on work [19, 6, 23, 11, 5, 9, 28, 31, 12].”
Efficient Off-Policy Meta-Reinforcement Learning via Probabilistic Context Variables (2019, methodology): “For supervised learning, Rusu et al. (2019); Gordon et al. (2019); Finn et al. (2018) adapt model predictions using probabilistic latent task variables inferred via amortized approximate inference.”
Meta-Learning in Neural Networks: A Survey (2020, methodology): “For instance, MAML can be improved by learning a learning rate per layer per step [76].”
Learning to Self-Train for Semi-Supervised Few-Shot Classification (2019, background): “They can be roughly divided into four categories: (1) data augmentation based methods [15, 29, 37, 38] generate data or features in a conditional way for few-shot classes; (2) metric learning methods [36, 31, 33] learn a similarity space of image features in which the classification should be effici”
Open-world Learning and Application to Product Classification (2018, methodology): “…example, it has been used to learn an optimizer [17], to learn network conﬁgurations [18], to learn initial and easy-to-tune weights for few-shot learning [19, 20], to learn a teacher model that can guide training sample selection [21] and to learn a domain training corpus for word embeddings [22].”
Generalizing from a Few Examples (2019, background): “eter of the model regression networks is learned by gradient descent like θ. 5.2.2 Refining in consideration of uncertainty. Learning with few-shot inevitably results in a model with high uncertainty [34]. Where the learned model can predict for new task with high confidence? Will the model be impro”
Meta-Transfer Learning Through Hard Tasks (2019, methodology): “Its task-agnostic property makes it possible to generalize to few-shot supervised/semi-supervised learning as well as unsupervised re-inforcement learning [7], [19], [20], [23]–[25].”
An Ensemble of Epoch-Wise Empirical Bayes for Few-Shot Learning (2019, methodology): “sses belongingto 4 super-classes, respectively. On all datasets, we consider the (5-class, 1-way) and (5- class, 5-way) classiﬁcation tasks. We use the same task sampling strategy as in related works [7, 8, 42]. The most related methods. MAML [7] is commonlyused as the baseline, and MAML++ [1] intro”
Reward Shaping via Meta-Learning (2019, methodology): “In light of this and recent works (Grant et al., 2018; Yoon et al., 2018; Finn et al., 2018), we call Φ( s ) the potential function prior .”