URL: https: //www.bilibili.com/video/av50747658/ (b station to find Chinese subtitles video)
the first week
I. Introduction
1.1 Welcome
1.2 What is Machine Learning
1.3 supervised learning
1.4 Unsupervised Learning
Second, univariate linear regression
2.1 model representation
2.2 cost function
2.3 I cost function intuitive understanding
2.4 cost function intuitive understanding II
Gradient descent 2.5
Gradient descent intuitive understanding 2.6
2.7 gradient descent of the linear regression
2.8 The following content
Third, Linear Algebra Review
3.1 Matrix and vector
3.2 addition and scalar multiplication
3.3 matrix-vector multiplication
3.4 Matrix Multiplication
3-5 matrix multiplication feature
(1) does not apply commutative matrix multiplication
(2) matrix multiplication is associative
(3) is a diagonal matrix (a11, a22, a33 ...) is equal to the matrix 1
3-6 inverse and transpose
(1) the inverse matrix
Only m * m matrix has an inverse matrix
I is the identity matrix
(2) transposed
aij of the matrix into aji
Week 2
Fourth, the multivariate linear regression
4-1 Multifunction
When the prediction with a plurality of rate parameters, need to use multiple linear regression, the vector is expressed as:
4-2 polyhydric gradient descent
1- Practice 4-3 wherein the scaling gradient descent
4-4 practice gradient descent learning rate 2-
4-5 features and polynomial regression
4-6 normal equation
4-7 normal equation and irreversibility (optional)
Five, Octave Tutorial
5-1 Basic Operations
= ~ Presentation logic is not equal to
5-2 Mobile Data
5-3 calculated data
5-4 draw data
5-5 Control statements: for while if statement
5-6 vector (did not quite understand)
The third week
Sixth, logistic regression
Category 6-1
Logistic regression algorithm (logistics regression) - logistic regression algorithm is a classification algorithm, which applies to the value of y is worth taking the discrete case.
Binary classification (Category 0,1)
6-2 assume statement
6-3 decision limits
6-4 cost function
6-5 to simplify the cost function and gradient descent
6-6 Advanced Optimization
End of this chapter need to achieve: write a function, it returns the value of the cost function, the gradient value, so this should be applied to logistic regression or even linear regression, you can also put these optimization algorithm for linear regression, you need to do enter the appropriate code is calculated such things here.
6-7 multivariate classification - many
y value is a plurality of classification value
Seven regularization
7-1 overfitting
What is the over-fitting
Regularization
7-2 cost function
7-3 linear regression regularization
7-4 logistic regression regularization
---------------------------------------
the fourth week
Eight, neural networks: representation
8-1 linear hypothesis
8-2 neurons in the brain
8-3 show model I
8-4 show model II
Examples 8-5 and intuitive understanding I
Examples 8-6 and intuitive understanding II
8-7 multivariate classification
9-1 cost function
9-2 back-propagation algorithm
9-3 understanding of back-propagation algorithm
9-4 Note: Expand the parameters
9-5 gradient detection
9-6 random initialization
9-7 grouped together
Unmanned 9-8
10-1 decide what to do next
10-2 assess the hypothesis
10-3 model selection and training, testing, validation set
60% 20% 20% divided into three sets (common)
10-4 diagnostic bias and variance
10-5 regularization and variance, deviation
10-6 Learning Curve
10-7 decide what to do next
11-1 determine the priority of execution
11-2 Error Analysis
11-3 asymmetry error evaluation Classification
11-4 precision and recall tradeoff Rate
11-5 Machine Learning Data
12-1 optimization goals
Spaced on the understanding intuitively 12-2
Mathematical Principles of 12-3 large margin classifier
12-4 kernel 1
12-5 Kernel 2
12-6 using SVM
13-1 unsupervised learning
Clustering
13-2 K-Means algorithm
13-3 optimization goals
13-4 random initialization
Select the number of clusters 13-5-
14-1 goal I: data compression
14-2 goal II: Visualization
14-3 Principal Component Analysis planning issues 1
14-4 Principal component analysis planning issues 2
Select the number of principal components 14-5
14-6 compression reproduce
PCA 14-7 Application recommendations
15-1 motivation problem
15-2 Gaussian distribution (normal distribution)
15-3 Algorithm
15-4 anomaly detection system development and evaluation
15-5 Anomaly Detection VS-supervised learning
When the positive samples is too small, large amount of negative samples when using the anomaly detection algorithm can learn from a negative sample a sufficient number of features
On the contrary, negative samples too little time, with supervised learning
15-6 select the function you want to use
15-7 multivariate Gaussian distribution
15-8 abnormality detection using a multivariate Gaussian distribution
16-1 planning issues
Recommended system
Content-based recommendation algorithm 16-2
16-3 collaborative filtering
16-4 collaborative filtering algorithm
Vectorization 16-5: low-rank matrix decomposition
16-6 implementation details: the mean standardized
17-1 studying large data sets
17-2 stochastic gradient descent
17-3 Mini-Batch gradient descent
17-4 Stochastic Gradient Descent
17-5 online learning
Mapping data in parallel to reduce 17-6
18-1 Problem Description and OCR.pipeline
Image Identification
18-2 sliding window
Using a sliding window to find the image detector of the pedestrian
18-3 acquiring large amounts of data and manual data
18-4 Ceiling analysis: pipeline of future work
19-1 Summary and thanks