Intro to Supervised Learning Notes(2)

Intro

1. High dimensional Feature

• what is the difference between low dimensional feature and high dimensional feature
• x = [x1,x2,…,xn] -> y
• infinite dimensional features
• select features based on the data

2. Regression vs Classification

• regression: if y is continuous variable,
  e.g., price prediction
• classification: if the label is a discrete variable
  e.g., the task of predicting the types of residence

3. Supervised learning in computer science

• Image Classification: x = raw pixels of the picture, and y = label
• Natural language processing

4. unsupervised learning

• only have data without labels
• goal is to find interesting structure in the data

5. Clustering & Other

• k-mean clustering, mixture of Gaussians
• clustering genes
• principal component analysis (tools used in
  LSA)
• word emeddings(Represent words by vectors),eg Word2vec
• clustering words with similar meanings

6. Reinforcement Learning

• learning to walk to the right
• the algorithms can collect data interactively
• method: try the strategy and collect feedbacks(Data Collections & Training),to improve the strategy based on the feedbacks

SL:Setup

• Linear Regression
• x -> y
• gradient descent

Linear Algebra

Probability

Gaussian Discriminant Analysis

Support Vector Machines. Kernels.

Evaluation Metrics

Reference

1. http://cs229.stanford.edu/notes2020spring/lecture1_slide.pdf
2. http://cs229.stanford.edu/notes2020spring/cs229-notes1.pdf
3. CS224N/CS231N
4. Identifying Regulatory Mechanisms using Individual Variation Reveals Key Role for Chromatin Modification.
5. Luo-Xu-Li-Tian-Darrell-M.’18