李宏毅 machine learning 笔记1 introduction

What is machine learning

machine learning ≈ looking for a function from data

Framework

Training

• step 1 Model (Define a set of function)
• step 2 Goodness of function
• step 2 Pick the best function

Learning Map

P.S. scenario n.情景，场景

Regression

• the out put of the target function is scalar

Classification

• Binary Classification
• Multi-class Classification

Classification - Deep Learning

• feature : the function has a hierarchical structure
• usage : image recognition, playing GO,

Supervised learning

• training data: input/output pair of target function
• function output = label
• hard to collect a large amount of labelled data

Semi-supervised learning

• training data included both labelled and unlabeled data

Transfer learning

• a part of data may not related to the task considered(can be labelled and unlabeled)

Unsupervised learning

• “learned without teacher”
• e.g. machine reading, machine drawing

Structured learning

• e.g. speech recognition, machine translation, face recognition

Reinforcement learning

• learning from critics(supervised one is learning from teacher)