机器学习基石 - Types of Learning

机器学习基石上 (Machine Learning Foundations)—Mathematical Foundations

Hsuan-Tien Lin, 林轩田，副教授 (Associate Professor)，资讯工程学系 (Computer Science and Information Engineering)

Types of Learning

Learning with Different Output Space

Binary Classification (二元分类)

Multiclass Classification 输出多个类别 （是非题变为选择题）

Regression (回归分析): deeply studied in statistics

Structured Learning 结构化学习 (huge multiclass classification problem)

Learning with Different Data Label ynyn

关于监督学习和非监督学习

Multiclass Classification 下的几种情况



Supervised Learning: every xnxn comes with corresponding ynyn

每个输入都知道对应的正确输出

Unsupervised: Learning without ynyn

clustering (聚类)

density estimation (密度分析)

outlier detection (离群点分析)

Semi-supervised: Learn with some ynyn

leverage unlabeled data to avoid ‘expensive’ labeling

标记的代价可能很大，只能做一部分

Reinforcement Learning (增强学习)

a very different but natural way of learning

惩罚错误的行为，奖励正确的行为

例子



learn with partial/implicit information (often sequentially) (一个一个的来学习)

Learning with Different Protocol f⇒(xn,yn)f⇒(xn,yn)

和机器的沟通方式

Batch Learning: 喂给机器一批一批的数据 (duck feeding)

监督学习、非监督学习

Online Learning: 一个一个的来 (sequentially)

PLA、增强学习

Active Learning: 主动的去学习 (ask questions)

query the ynyn of the chosen xnxn

improve hypothesis with fewer labels (hopefully) by asking questions strategically

机器无法识别时，让机器自己来问，需要标记的可能就少一些

总结

Learning with Different Input Space XX

Concrete Features

each dimension of X⊆RdX⊆Rd represents sophisticated physical meaning

具体的东西，可以计算，预先有人类智慧的加工

the easy ones for ML

Raw Features

更为抽象，包含很多细节 simple physical meaning

like image pixels, speech signal, etc.

often need human or machines to convert to concrete ones

feature engineer (特征工程) 提取出哪些特征给机器比较好

Abstract Features

需要机器自己去学到特征

no physical meaning

need feature conversion/extraction/construction

例如只有一些编号的对应关系

思考题