arXiv research

A new meta-learning framework that assigns weights to source tasks based on target samples.

Meta-learning improves neural networks by adapting learning algorithms.

New PAC-Bayesian framework for flexible meta-learning.

Meta-learning helps models learn quickly from few samples.

Meta-learning balances task-specific modeling and optimization complexity.

Meta-learning algorithm improves AI efficiency by teaching itself.

Meta learning can adapt fast but is vulnerable to adversarial attacks.

Meta-learning algorithms are studied for their data efficiency in personalized federated learning.

The paper provides fast, sample-efficient algorithms for learning common features in multi-task linear regression.

This paper improves meta-learning by developing new PAC-Bayes bounds.

Meta-learning enables a model to learn from very limited data to undertake a new task. In this paper, we study the general meta-learning with adversarial samples. We present a meta-learning algorithm, ADML (ADversarial Meta-Learner), which leverages clean and adversarial samples to optimize the initialization of a lear…

Meta-learning framework uses task similarity through nonparametric kernel regression.

Meta-learning algorithms use past experience to learn to quickly solve new tasks. In the context of reinforcement learning, meta-learning algorithms acquire reinforcement learning procedures to solve new problems more efficiently by utilizing experience from prior tasks. The performance of meta-learning algorithms depe…

Self-referential meta learning avoids explicit optimization by modifying itself.

Meta-learning is a promising strategy for learning to efficiently learn within new tasks, using data gathered from a distribution of tasks. However, the meta-learning literature thus far has focused on the task segmented setting, where at train-time, offline data is assumed to be split according to the underlying task,…

learn2learn simplifies meta-learning research by providing a library and standardized interfaces.

Probabilistic active meta-learning improves data efficiency in robotics.

Meta-learning algorithms produce feature extractors which achieve state-of-the-art performance on few-shot classification. While the literature is rich with meta-learning methods, little is known about why the resulting feature extractors perform so well. We develop a better understanding of the underlying mechanics of…

New analysis improves generalization bounds for meta-learning.

Meta learning with information theory and Gaussian processes.

Many IoT applications at the network edge demand intelligent decisions in a real-time manner. The edge device alone, however, often cannot achieve real-time edge intelligence due to its constrained computing resources and limited local data. To tackle these challenges, we propose a platform-aided collaborative learning…

New PAC-Bayes meta-learning method improves few-shot learning accuracy and calibration.

TaskNorm improves meta-learning performance by rethinking batch normalization.

Meta-learning bounds derived using PAC-Bayes theory for improved generalization.

New meta-learning framework for minimizing simple regret in bandits.

There is no free lunch, no single learning algorithm that will outperform other algorithms on all data. In practice different approaches are tried and the best algorithm selected. An alternative solution is to build new algorithms on demand by creating a framework that accommodates many algorithms. The best combination…

We extend first-order model agnostic meta-learning algorithms (including FOMAML and Reptile) to image segmentation, present a novel neural network architecture built for fast learning which we call EfficientLab, and leverage a formal definition of the test error of meta-learning algorithms to decrease error on out of d…

Algorithm reduces regret in non-stationary bandits and meta-learning with optimal arms.

Unified framework improves meta-learning generalization bounds.

Meta-learning priors improves safe Bayesian optimization.

Study MAML's generalization in varying tasks, proving bounds on error.

BOML unifies meta-learning methods into a common bilevel optimization framework.

New algorithm for online meta-learning with task boundary detection.

A meta-learning approach for efficient algorithm selection in budget-limited scenarios.

This paper bounds meta-generalization gap using information theory.

Meta-learning curiosity algorithms improves exploration across various tasks.

Meta-learning method for accurate classifier from noisy annotators' data.

VSML unifies meta learning concepts and enables simple backpropagation.

Improves meta-learning efficiency with mixed-mode differentiation.

We introduce a new, rigorously-formulated Bayesian meta-learning algorithm that learns a probability distribution of model parameter prior for few-shot learning. The proposed algorithm employs a gradient-based variational inference to infer the posterior of model parameters to a new task. Our algorithm can be applied t…

The ability to learn new concepts with small amounts of data is a critical aspect of intelligence that has proven challenging for deep learning methods. Meta-learning has emerged as a promising technique for leveraging data from previous tasks to enable efficient learning of new tasks. However, most meta-learning algor…

New PAC-Bayesian bounds explain few-shot learning performance gaps.

The constant introduction of standardized benchmarks in the literature has helped accelerating the recent advances in meta-learning research. They offer a way to get a fair comparison between different algorithms, and the wide range of datasets available allows full control over the complexity of this evaluation. Howev…

Meta-learning performance improved by keeping support sets fixed.

Meta-learning improves performance in stochastic linear bandits.

Proposes a method to evaluate meta-learning performance based on task similarity.

Meta-learning is increasingly used to support the recommendation of machine learning algorithms and their configurations. Such recommendations are made based on meta-data, consisting of performance evaluations of algorithms on prior datasets, as well as characterizations of these datasets. These characterizations, also…

VMGP extends Gaussian processes for Bayesian meta-learning, improving uncertainty prediction.