Artificial Intelligence
原型网络
Prototypical NetworksA class of metric learning algorithms that classify data points by computing the distance between a data point and the class prototype (the mean of the support set). / 一种度量学习算法类别，通过计算数据点与类别原型（支持集的均值）之间的距离来对数据点进行分类。
- Knowledge type
- 数学理论
- First proposed or discovered
- 2017
- Proposer
- Snell, Swersky, Zemel
- Creator
- 农场主
- Last updated
- 2026-10-26 16:01:34
Sub-knowledge points
Few-shot learningMetric learningPrototype computation
Higher-order knowledge points
Meta-learningSupport Vector MachinesDeep Metric Learning
Main parameters and performance indicators
k: Number of classesn: Number of shots per classembedding_dim: Dimension of the feature space
Open problems
Handling extremely small sample sizesDealing with class imbalanceTask-adaptive feature learning
Related literature
ranked by community starsTask-Adaptive Multi-Prototype Prototypical Networks for keratoconus diagnosis under unbalanced small samplesOpen source
Related products and links
No item recorded yet
Creation and modification trail
1 recordsRecord title: Prototypical Networks.
Node update feedback
review required