neuralnet_function_approximation.Rmd

---
title: "neuralnet function approximation"
author: "PavanMirla"
date: '2017-02-12'
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

## Neural Net to approximate y=x^2



```{r neuralnet}
library(neuralnet)
set.seed(2016)
attribute <- as.data.frame(sample(seq(-2,2,length = 50), replace = FALSE),ncol = 1)
response <- attribute^2
data <- cbind(attribute, response)
colnames(data) <- c("attribute","response")
plot(data)

```

## Fit a DNN with two hidden layers each containing three neurons

```{r}
fit <- neuralnet(response ~ attribute, data = data, hidden = c(3,3), threshold = .01)
```

## How good the model really is at approximating a function

```{r}


test_data <- as.matrix(sample(seq(-2,2,length = 10), replace = FALSE), ncol = 1)
#predict using the compute function

pred <- compute(fit, test_data)

result <-  cbind(test_data, pred$net.result, test_data^2)
colnames(result) <- c("Attribute", "Prediction", "Actual")
plot(result)

# Attribute is Predictor



```