1. Meta

描述数据的数据

2. Meta Learning

• Learning Algorithm: Learn from historical data and make predictions given new examples of data
• Meta Learning Algorithm: Learn the output of learning algorithms and make a prediction given predictions made by other models

3. Few-shot Learning

3.1. Support Set vs. Training Set

3.2. Few-shot Learning vs. Supervised Learning

• Supervised Learning
  • query样本类别在训练集中出现
• Few-shot Learning
  • query样本来自未知类别 $$\mathcal{T} \sim P(\mathcal{T}) \\ \mathcal{T} = (S_{\mathcal{T}}, Q_{\mathcal{T}}) \\ S_{\mathcal{T}} = S_{\mathcal{T}}^s \cup S_{\mathcal{T}}^u \\ Q_{\mathcal{T}} = Q_{\mathcal{T}}^s \cup Q_{\mathcal{T}}^u \\ $$