3.1: Introduction to Session 3 – What is Machine Learning?

Machine learning overview

Common Algorithm （Machine Learning Recipe）：

• KNN [k近傍法]

• SVM
• *ANN

Types of learning

• Supervised learning
  • Tech the system with the input and outputs
  • Have training data
  • Have test data (the data excludes from training data)
  • Have Samples (Unknown data)
• Unsupervised learning
  • Data that know nothing about *
    • e.g. songs patterns
  • Identify the commonalities in the data
• *Reinforcement Learning
  • Reward based

3.2: Linear Regression with Ordinary Least Squares Part 1 – Intelligence and Learning

Linear Regression「線形回帰せんけいかいき」

A linear approach to modelling the relationship between a scalar response

Ordinary Least Squares (OLS)「最小二乗法さいしょうにじょうほう」

a type of linear least squares methods for estimating the unknown parameters in a linear regression model.

Linear Model formula:

formula to find the slop:

(where  and  are the average)

issue with different data sets

When to use linear regression?

ref link: https://kwichmann.github.io/ml_sandbox/linear_regression_diagnostics/

• Residual plot should be random

3.4 Linear Regression with Gradient Descent

Gradient Descent 「最急降下法さいきゅうこうかほう」

Gradient descent is a first-order iterative optimization algorithm for finding the minimum of a function.

Required Math

• Calculus
  • derivative
  • partical derivative

Basic steps to implement the gradient descent

1. Obtain the results from the correct result sets
2. Obtain the evaluated results from the program
3. Calculate the error (where the error is the bias from program result to the correct results)
4. Adjusted the program parameter according to the learning rate (here is where the calculus comes in)