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

Meta-learning models transfer the knowledge acquired from previous tasks toquickly learn new ones. They are trained on benchmarks with a fixed number ofdata points per task. This number is usually arbitrary and it is unknown how itaffects performance at testing. Since labelling of data is expensive, findingthe optimal allocation of labels across training tasks may reduce costs. Givena fixed budget of labels, should we use a small number of highly labelledtasks, or many tasks with few labels each? Should we allocate more labels tosome tasks and less to others? We show that: 1) If tasks are homogeneous, thereis a uniform optimal allocation, whereby all tasks get the same amount of data;2) At fixed budget, there is a trade-off between number of tasks and number ofdata points per task, with a unique solution for the optimum; 3) When trainedseparately, harder task should get more data, at the cost of a smaller numberof tasks; 4) When training on a mixture of easy and hard tasks, more datashould be allocated to easy tasks. Interestingly, Neuroscience experiments haveshown that human visual skills also transfer better from easy tasks. We provethese results mathematically on mixed linear regression, and we showempirically that the same results hold for few-shot image classification onCIFAR-FS and mini-ImageNet. Our results provide guidance for allocating labelsacross tasks when collecting data for meta-learning.

How to distribute data across tasks for meta-learning?