일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | ||
6 | 7 | 8 | 9 | 10 | 11 | 12 |
13 | 14 | 15 | 16 | 17 | 18 | 19 |
20 | 21 | 22 | 23 | 24 | 25 | 26 |
27 | 28 | 29 | 30 |
- animation retargeting
- NLP
- ddpm
- cv
- UE5
- WBP
- dl
- motion matching
- Generative Model
- WinAPI
- RNN
- 모션매칭
- BERT
- multimodal
- ue5.4
- CNN
- GAN
- userwidget
- Stat110
- 디퓨전모델
- 오블완
- 폰트생성
- 딥러닝
- Diffusion
- Few-shot generation
- 언리얼엔진
- Font Generation
- 생성모델
- deep learning
- Unreal Engine
- Today
- Total
Deeper Learning
[Statistics 110] Lecture 12: Discrete vs. Continuous, the Uniform 본문
[Statistics 110] Lecture 12: Discrete vs. Continuous, the Uniform
Dlaiml 2022. 12. 10. 12:33Probability Density Function(PDF)
정의
확률변수 X가 모든 a,b에 대해 P(a≤X≤b)=∫baf(x)dx 를 만족시키면, X는 확률밀도함수(PDF) fX(x)를 가짐
a=b라면 적분값은 0 = 특정 값을 가질 확률은 0
pmf와 마찬가지로 음수가 아니며, 모든 구간에 대해 적분하면 1이 되어야 함
Density의 의미

CDF와 PDF의 관계
FX(x)=P(X≤x)=∫x−\infinf(t)dt
pdf fX(x)=cdf F′(x)
기댓값
E(X)=∫\infin−\infinxfX(x)dx
분산, 표준편차
Var(x)=E((X−E(X))2)=E(X2)−(E(X))2
SD(X)=√Var(X)
Uniform Distribution(균등분포)
Unif(a,b)는 [a,b] 범위에서 랜덤 포인트의 위치
확률이 범위와 비례하는 분포
f(x) = c (if a ≤ x ≤b) otherwise 0
(a~b로 적분하면 1이 나와야 한다 → c = 1 / (b-a)
cdf
F(x) = 0 (if x < a), 1 (if x > b), (x-a) / (b-a) (if a ≤ x ≤ b)
Expected value
E(X)=∫baxb−adx=a+b2
Variance
X의 제곱의 기댓값을 알아야 계산이 가능, Y=X2 라고 하면
E(X2)=E(Y)
E(X)=∫\infin−\infinxfX(x)dx
E(X2)=∫\infin−\infinx2fX(x)dx
X2의 기댓값을 구하지 않고 X가 따르는 pdf를 그대로 사용하고 앞의 값만 X2으로 바꾼 세 번째 식은 성립한다.
이를 Law of the unconscious statistician(LOTUS, 무의식적인 통계학자의 법칙)이라고 부른다.
일반화한 수식은 아래와 같다. (임의의 함수 g)
E(g(X))=∫\infin−\infing(x)fX(x)dx,∑xg(x)P(X=x)
u가 0~1 범위의 uniform distribution을 따르는 확률변수라고 하자
E(u)=1/2E(u2)=∫10u2fu(u)du=1/3Var(u)=E(u2)−E(u)2=1/12
Universality of the uniform distribution
u가 0~1 범위의 uniform distribution을 따르는 확률변수라고 하자
F는 CDF (증가하는 연속함수로 가정, 역함수가 존재)
X를 F−1(u)라고 하면 확률변수 X는 F를 따른다
proof)
증가하는 연속함수라 부등식 양변에 F를 적용할 수 있음
P(X≤x)=P(F−1(u)≤x)=P(F(F−1(u))≤F(x))=P(u≤F(x))

u는 0~1 범위 내에 있기 때문에
범위 내 일정 지점까지의 확률은 해당 길이와 같다
- u가 0.5에 위치한 점보다 왼쪽에 있을 확률은 0.5
따라서 P(u≤F(x))=F(x)
Reference
[0] https://www.youtube.com/playlist?list=PL2SOU6wwxB0uwwH80KTQ6ht66KWxbzTIo
Statistics 110: Probability
Statistics 110 (Probability) has been taught at Harvard University by Joe Blitzstein (Professor of the Practice in Statistics, Harvard University) each year ...
www.youtube.com
'Statistics & Math > Statistics 110: Probability' 카테고리의 다른 글
[Statistics 110] Lecture 14: Location, Scale, and LOTUS (0) | 2022.12.17 |
---|---|
[Statistics 110] Lecture 13: Normal Distribution (0) | 2022.12.10 |
[Statistics 110] Lecture 11: The Poisson distribution (0) | 2022.12.10 |
[Statistics 110] Lecture 10: Expectation Continued (0) | 2022.12.10 |
[Statistics 110] Lecture 9: Expectation, Indicator Random Variables, Linearity (0) | 2022.12.08 |