Week 1

Andrew Ng 교수님의 Machine Learning 강좌 주요 키워드 및 내용입니다.

Introduction

Definition of Machine Learning

• Learn without being explicitly programmed
• For task T, with lots of experience E, performance P would be better

Categorization of Machine Learning

• Supervised Learning and Unsupervised Learning
• Regression and Classification
• Cocktail Party Problem

Linear Regression with One Variable

Univariate Linear Regression

• Hypothesis: $h_θ(x)=θ_0+θ_1x$
• Cost Function: $J(θ)=\frac{1}{2m} \Sigma_{i=1}^{m} (h_θ(x_i)-y_i)^2=\frac{1}{2m} \Sigma_{i=1}^{m}(θ_0+θ_1x_i-y_i)^2$
• Objectives: min_θJ(θ)

Gradient Descent Algorithm

• $θ_0 :=θ_0-α\frac{δ}{δθ_0}J(θ)=θ_0-α\frac{1}{m}\Sigma_{i=1}^{m}(h_θ(x_i)-y_i)$
• $θ_1 :=θ_1-α\frac{δ}{δθ_1}J(θ)=θ_1-α\frac{1}{m}\Sigma_{i=1}^{m}(h_θ(x_i)-y_i)x$

Convex Function

Stochastic Gradient Descent Algorithm

 

Linear Algebra Review

Matrix/Vector Calculation

Identity Matrix

Matrix Inversion