# Machine Learning

Books and Courses:

Data:

To research:

Tips:

## Coursera Course

• m = number of training examples
• x = input variable
• y = output/target variable

### Notation

Colon-equals is for assignment:

 a := b 

An equals sign is truth assertion:

 a = b 

Greek lowercase alpha is the learning rate.

m is the number of rows of data.

### Linear Regression

Hypothesis: $$h\theta(x)=\theta_{0}+\theta_{1}x$$

Parameters: $$\theta_{0}$$ and $$\theta_{1}$$.

Cost function: $$J(\theta_{0},\theta_{1})=\frac{1}{2m}\sum_{i=1}^m(h\theta(x^{(i)})-y^{(i)})^{2}$$

Goal: $$\underset{\theta_0,\theta_1}{\text{minimize}}~J(\theta_{0},\theta_{1})$$

Gradient descent algorithm: $$\theta_{j}:=\theta_{j}-\alpha\frac{\partial}{\partial\theta_{j}}J(\theta_{0},\theta_{1})$$

#### Simultaneous Update

Be sure to use simultaneous update. Example:

$$temp0:=\theta_{0}-\alpha\frac{\partial}{\partial\theta_{0}}J(\theta_{0},\theta_{1})\\temp1:=\theta_{1}-\alpha\frac{\partial}{\partial\theta_{1}}J(\theta_{0},\theta_{1})\\\theta_{0}:=temp0\\\theta_{1}:=temp1$$

More notes from the course:

1. Basics
• Supervised Learning
• Regression problems -- "trying to predict results within a *continuous* output" e.g., predicting price based on house size
• Classification problems -- "trying to map input variable to some *continuous* function" e.g., whether a house sells for more or less than the asking price (discrete categories)
• Unsupervised Learning
• Clustering articles into groups based on similarity
• Associative -- like a doctor associating possible illnesses based on what has been seen in previous patients
1. Linear Regression, One Variable
• Linear regression -> continuous expected result function
• Univariate linear regression -> for single output from single input value
• Hypothesis function: $$h\theta(x)=\theta_{0}+\theta_{1}x$$
• Cost function: $$J(\theta_{0},\theta_{1})=\frac{1}{2m}\sum_{i=1}^m(h\theta(x^{(i)})-y^{(i)})^{2}$$ -- measures the accuracy of the hypothesis function
• Goal: $$\underset{\theta_0,\theta_1}{\text{minimize}}~J(\theta_{0},\theta_{1})$$