1. ML的基本分类:
1.1 不同的\(y\)类型:
- 二元分类:
- 多元分类:
- 回归:
- 结构学习:
1.2 是否有\(y\):
有监督(supervised):
Coin Recognition:every \(x_n\) comes with corresponding \(y_n\)无监督(unsupervised):
半监督(semi-supervised):
Semi-supervised: Coin Recognition with Some yn
强化学习(reinforcement):
1.3 不同的学习策略(用户和机器的交互方式):
批处理(batch learning):
batch supervised multiclass classification: learn from all known data- 在线学习(online learning):
尤其适用于结合流式计算,即数据是逐个产生,串行输入,无法batch
【注】:
- 联想PLA的最简单实现版本,就是online-learning:每次发现一个点犯错,就立刻修改\(w\),知错就改
- reinforcement learning is often done online :因为需要逐步修正model
- online hypothesis ‘improves’ through receiving data instances sequentially
- Active(主动学习):
improve hypothesis with fewer labels (hopefully) by asking questions strategically
1.4 不同的输入空间\(\mathcal{X}\):
concrete faeture:
concrete features: each dimension of \(X ⊆ R^d\) represents 'sophisticated' physical meaningRaw Features:
such as Digit Recognition Problem,raw input space like this,- Other Problems with Raw Features: image pixels, speech signal, etc.
- raw features: often need human or machines to convert to concrete ones.
- Abstract Features:
abstract: again need ‘feature conversion/extraction/construction’