Machine Learning Overview

What is machine learning? What is it good for?

1. What is machine learning?
   
   1. Knowing the equation and x before, find y. Machine learning is knowing x and y, 求方程.
2. What is the use?
   
   1. Machine learning can be used to solve problems such as regression prediction, classification, and clustering.
3. How to learn machine learning
   
   1. https://www.zhihu.com/question/51039416/ How can ordinary programmers learn the knowledge of artificial intelligence correctly?
   2. Getting started with CS229, you can take a look at "Python Machine Learning"
   3. In the advanced stage, you can practice on kaggle. You can take a look at Zhou Zhihua's "Machine Learning"
   4. Do deep learning.
   5. In-depth research, subscribe to arvix, pay attention to the top meeting, and start keeping up with the industry!
4. Machine Learning Operation Routines
   
   1. First of all, of course, the problem is transformed into a mathematical language. For example, a linear regression formula is used to represent the problem. (various models are involved here)
   2. Then get a formula that can be optimized,$loss function = 预测函数 -真实值$.
   3. Then continue to optimize (fit), which involves many optimization methods.
      
      1. gradient descent. Local optimum.
         
         1. Least Mean Square Least Mean Square Algorithm
      2. Newton's Method. Global Optimum.
      3. Exponential family.
      4. Generalized Linear Model.
   4. Of course, overfitting is not good, so we introduce some methods to prevent generalization.
      
      1. Regularization.

Concept tree/mindmap/ directory

1. Machine learning algorithms can be roughly divided into three types
   
   1. Supervised learning. If there is input data 标签, it is supervised learning.
      
      1. regression algorithm.
      2. classification algorithm.
   2. Unsupervised learning.
      
      1. clustering
      2. Dimensionality reduction
   3. Reinforcement learning. Achieving the best policy through constant trial and error.
      
      1. Enhanced learning. Through continuous trial and error, the model is corrected. Suitable for various games;
      2. 试错学习, there is no direct guidance, and the optimal strategy needs to be obtained through constant trial and error.
      3. 延迟回报, all rewards are known only after the end of the game, so the final rewards need to go back to the previous state.
2. Two types of machine learning algorithms
   
   1. Generative learning algorithms.
      
      1. Gaussian discriminant analysis.
      2. Naive Bayes.
      3. HMM.
   2. Discriminative learning algorithms.
      
      1. SVM.
      2. Logistic Regression.
   3. Personal summary
      
      1. https://stackoverflow.com/a/879591 comes to the root, generative==joint probability & discriminative==
      2. The discriminant is to draw a line, and the line can absolutely distinguish the categories; the generative is to generate a model (or a range/a circle) through the known data, and the subsequent judgment falls in this circle is this category.
      3. What is the difference between "decision model" and "generative model" in machine learning? https://www.zhihu.com/question/20446337

Authoritative information

1. teaching material
   
   1. Machine Learning Zhou Zhihua
      
      1. It is recommended to use the watermelon book as a reference book rather than a main reading book.
      2. Due to the limitation of space, the watermelon book covers a lot of content but cannot be viewed in detail. For beginners, the actual reading is very large.
      3. This book is more suitable for use as a school textbook or self-study by intermediate readers. It is slightly more difficult to learn this book when getting started.
   2. Machine Learning Foundation Machine Learning Cornerstone
   3. Machine Learning Techniques
2. Undergraduate courses
   
   1. http://cs229.stanford.edu/
   2. http://www.cs.cmu.edu/~tom/10701_sp11/lectures.shtml
   3. http://www.cs.cmu.edu/~ninamf/courses/601sp15/lectures.shtml I don't know the difference between these two CMU courses.
   4. http://www.cs.cmu.edu/~aarti/Class/10701_Spring14/
   5. ML class at the Max Planck Institute for Intelligent Systems.
   6. https://work.caltech.edu/telecourse Yaser is Lin Xuantian's teacher.
   7. Metacademy is a great learning material for machine learning and related math.
3. PhD level courses
   
   1. CMU 10-702(Statistical Machine Learning)
   2. CMU 10-715(Advanced Introduction to Machine Learning)
4. MIT 6.S099: Artificial General Intelligence by Lex Fridman
5. https://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-867-machine-learning-fall-2006/syllabus/