Machine Learning includes Supervised Learning, Unsupervised Learning, Reinforcement Learning.


Supervised Learning:

Given a dataset that contains n samples (x1, y1) (x2, y2)...
Find relation between features and output
and then do prediction

Regression VS Classification

Regression:
if y is a continuous variable (eg. price predicion)

Classification:
label is a discrete variable

Unsupervised Learning:

Dataset has no labels

Clustering:

Gaussian mixture model, 
k-means, 
PCA

Use for CV:

Object detection
Objection localization
Image Classification

Use for NLP:

Machine translation

Reinforcement Learning:

Collect data interactively
Try strategy and collect feedbacks
Improve the strategy based on the feedbacks

E.g.:

AI play video games
Alpha Go
MuZero (Dec 2020), Deepmind, more general, play all games.

Word “Embeddings”:
- Represent words by vectors
- Represent relation by direction

Senticap:
- Describing Images with Sentiments

SemStyle:
- Learning to Caption from Romantic Novels