3dgaussianmixturemodel.rmd

---
title: "Gaussian Mixture Model"
author: "Daniel Frederico Lins Leite"
date: "21 April 2017"
output: ioslides_presentation
---

```{r setup, include=FALSE, warning=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(plotly)
library('MASS')
```

## Gaussian Mixture Models

- First we will generate 2000 points from two different multivariate Gaussian distributions.
- Then we will generate a 2d histogram of this data and render in 3D the possible approximation of the Gaussian, a result that we could have achieved from a Gaussian Mixture Model.

## Code

```{r, warning=FALSE, }
simulation.g1 <- mvrnorm(n = 1000, c(0,0), diag(2))
simulation.g2 <- mvrnorm(n = 1000, c(-3,-3), diag(c(0.3,10)))
simulation <- rbind(simulation.g1, simulation.g2)
qtd <- kde2d(simulation[,1], simulation[,2])
```

## Plot

```{r, warning=FALSE}
plot_ly(x = qtd$x, y = qtd$y, z = qtd$z) %>% add_surface()
```