English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Existing gradient-based meta-learning approaches to few-shot learning assumethat all tasks have the same input feature space. However, in the real worldscenarios, there are many cases that the input structures of tasks can bedifferent, that is, different tasks may vary in the number of input modalitiesor data types. Existing meta-learners cannot handle the heterogeneous taskdistribution (HTD) as there is not only global meta-knowledge shared acrosstasks but also type-specific knowledge that distinguishes each type of tasks.To deal with task heterogeneity and promote fast within-task adaptions for eachtype of tasks, in this paper, we propose HetMAML, a task-heterogeneousmodel-agnostic meta-learning framework, which can capture both thetype-specific and globally shared knowledge and can achieve the balance betweenknowledge customization and generalization. Specifically, we design amulti-channel backbone module that encodes the input of each type of tasks intothe same length sequence of modality-specific embeddings. Then, we propose atask-aware iterative feature aggregation network which can automatically takeinto account the context of task-specific input structures and adaptivelyproject the heterogeneous input spaces to the same lower-dimensional embeddingspace of concepts. Our experiments on six task-heterogeneous datasetsdemonstrate that HetMAML successfully leverages type-specific and globallyshared meta-parameters for heterogeneous tasks and achieves fast within-taskadaptions for each type of tasks.

HetMAML: Task-Heterogeneous Model-Agnostic Meta-Learning for Few-Shot Learning Across Modalities