DeepLearning/인공지능 개론
인공지능 및 기계학습 개론 2(8.4~8.5) 정리
8.4 Multivariate Gaussian Distribution Multivariate Gaussian Distribution Probability density function of the Gaussian distribution $N(x |\mu, \sigma^2) = \frac{1}{\sqrt{2\pi\sigma^2}}\exp(-\frac{1}{\sigma^2}(x - \mu)^2)$ $N(x |\mu, \sum) = \frac{1}{(2\pi)^{D/2}}\frac{1}{|\sum|^{1/2}}\exp(-\frac{1}{2}(x-\mu)^T\sum^{-1}(x-\mu))$ → 여기서 $\sum$은 covariance matrix가 된다. → 이제 우리는 Multivariate Guassian ..
DeepLearning/인공지능 개론
인공지능 및 기계학습 개론 2(8.1~8.3) 정리
8.1 ~ 2 K-Means Algorithm Unsupervised Learning You don’t know the true value, and you cannot provide examples of the true value. Cases, such as Discovering clusters Discovering latent factors Discovering graph strcutures Clustering or filtering or completing of Finding the representative topic words from text data Finding the latent image from facial data Completing the incomplete matrix of pro..
DeepLearning/인공지능 개론
인공지능 및 기계학습 개론 2(7.7~7.9) 정리
7.7 Variable Elimination Marginalization and Elimination Computing joint probabilities is a key How to compute them? Many, many, many multiplications and summations $P(a=ture, b=ture, mc=ture) = \sum_{jc}\sum_{E}P(a,b,E,JC, mc) \= \sum_{JC}\sum_E P(JC|a)P(mc|a)P(a|b,E)P(E)P(b)$ 기억이 안나서 메모하자면, marginalization하는 과정은 쉽게 생각해서 hidden variable의 모든 상태(?)대해서 summation하면 되고, $P(a,b,E,JC, mc)$가 아래처럼 바뀐 이유..
DeepLearning/인공지능 개론
인공지능 및 기계학습 개론 2(7.4~7.6) 정리
7.4 Bayes Ball Algorithm Typical Local Structures Common parent Fixing “alarm” decouples “JohnCalls” and “MarryCalls” $J \bot M|A$ → A에 대한 정보가 있을 때, $J$와 $M$은 independent하다. $P(J, M|A) = P(J|A)P(M|A)$ Cascading Fixing “alarm” decouples “Buglary” and “MaryCalls” $B\bot M|A$ → A에 대한 정보가 있을 때 $B$와 $M$은 independent하다. $P(M|B,A) = P(M|A)$ V-Structure Fixing “alarm” couples “Buglary” and “Earthquake” ..
