Introduction
temporary created in order to be referenced by other posts.
Perceptron and MLP
Posted on
|
In
machine_learning
EM Algorithm and Gaussian Mixture Model
Posted on
|
In
machine_learning
Introduction to EM
Previously I talked about Naive Bayes Classifier. In that case, all variables are observable. However, sometimes there could be some latent variables in some models. EM(Expection-Maximization) algorithm could resolve those models containing latent variables.
Bayes Classifier
Posted on
|
In
machine_learning
General Idea
Generally speaking, to solve a classification problem, what we need to do is to get the optimal estimation on the posterior:
$$h(x) := \underset{c_k}{\arg \max} , P(Y=c_k|X=x) \tag{1}\label{eq:1}$$
There are 2 general methods to do this. First one is directly modeling the posterior based on training data and to do prediction based on that. This is called Discriminative Model, e.g., BP, LDA, SVM, etc. Another type is generative Models, which calculates $P(Y=c_k|X=x)$ using:
$$P(Y=c_k|X=x) := \frac {P(x,c_k)}{P(x)}\tag{2}\label{eq:2}$$
Naive Bayes Classifier is the second one.
Backpropagation Algorithm
Posted on
|
In
machine_learning