Optimization (최적화) in Statistics

Prof. Daewon Yang (Chungnam National University)

Optimization (최적화)

Graph of a given by z = f(x, y) = −(x^2 + y^2) + 4. The global maximum at (x, y, z) = (0, 0, 4) is indicated by a blue dot.

Section 1. Optimization Problem

1.1 Optimization Problem의 수학적 정의.

constraint

plot(-100,100, xlim=c(-5,5), ylim=c(-5,5), xlab="X1", ylab="X2", main="A")
segments(x0=1, y0=1, x1=10, y1=1, lwd=2, col=2)

1.2 Minimum and Maximum

Graph of cos(3\pi x)/x for 0.1 \le x \le 1.1

Global minimum (maximum)

local minimum (maximum)

Notation

1.3 Other Deifinitions

Continuous

Differentiable

open & closed & boundary

bounded

compact

Gradient

The gradient of the function f(x,y) = −(cos^2x + cos^2y)^2

Critical point

Saddle point

A saddle point (in red) on the graph of z = x^2 − y^2

1.4 Global minimum 계산 방법

Property

1.5 Convex optimization

Section 2. Optimization의 계산 테크닉

2.1 Gradient descent

2.2 Newton’s method

2.3 Lagrange multiplier

2.4 Machine learning에서의 optimization 테크닉

Section 3. 통계학 분야에서의 Optimization Problem의 활용.

3.1 Linear regression

3.2 Generalized linear model

3.3 Lasso and Ridge regression

3.4 EM algorithm