data-driven algorithmslearning to learnmachine learning pipelinesmeta-learningneural architectures
Figures from the paper
Abstract (AI)
Meta-learning, or learning to learn , is the science of systematically observing how different machine learning approaches perform on a wide range of learning tasks, and then learning from this experience, or meta-data , to learn new tasks much faster than otherwise possible. Not only does this dramatically speed up and improve the design of machine learning pipelines or neural architectures, it also allows us to replace hand-engineered algorithms with novel approaches learned in a data-driven way. In this chapter, we provide an overview of the state of the art in this fascinating and continuously evolving field.
Key Findings
1
By learning from prior task experience, meta-learning can enable new tasks to be learned substantially faster than with conventional approaches.
2
Data-driven meta-learning can replace manually engineered algorithms with novel approaches learned from experience.
3
Meta-learning can improve the design of machine-learning pipelines and neural architectures.
4
Meta-learning systematically studies machine-learning performance across diverse tasks and uses the resulting meta-data to accelerate learning on new tasks.
5
The field encompasses a broad and continuously evolving state of the art rather than a single fixed methodology.
Research Object
machine learning approaches and learning tasks
Research Subject
systematic learning from meta-data to improve and accelerate learning new tasks and the design of machine learning pipelines or neural architectures
Publication Details
Publication Date
2019-01-01
Journal
Publisher
ISSN
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest