R stats code for building NN and looping over hidden layer node density for animated gif output. NN code attributed to David Selby; http://selbydavid.com/2018/01/09/neural-network/

Beach not Beach NN Loop.r

########################################################################
### Bespoke Neural Network R code attributed to: David Selby
### From blog post: http://selbydavid.com/2018/01/09/neural-network/
### Adapted here for making animated GIF of node density
### output gifs compiled at gifmaker.me for final output
### output tweeted here:
### https://twitter.com/Md_Harris/status/951257342418608128
########################################################################

two_spirals <- function(N = 200,
                        radians = 3*pi,
                        theta0 = pi/2,
                        labels = 0:1) {
  N1 <- floor(N / 2)
  N2 <- N - N1
  
  theta <- theta0 + runif(N1) * radians
  spiral1 <- cbind(-theta * cos(theta) + runif(N1),
                   theta * sin(theta) + runif(N1))
  spiral2 <- cbind(theta * cos(theta) + runif(N2),
                   -theta * sin(theta) + runif(N2))
  
  points <- rbind(spiral1, spiral2)
  classes <- c(rep(0, N1), rep(1, N2))
  
  data.frame(x1 = points[, 1],
             x2 = points[, 2],
             class = factor(classes, labels = labels))
}
feedforward <- function(x, w1, w2) {
  z1 <- cbind(1, x) %*% w1
  h <- sigmoid(z1)
  z2 <- cbind(1, h) %*% w2
  list(output = sigmoid(z2), h = h)
}

sigmoid <- function(x) 1 / (1 + exp(-x))

backpropagate <- function(x, y, y_hat, w1, w2, h, learn_rate) {
  dw2 <- t(cbind(1, h)) %*% (y_hat - y)
  dh  <- (y_hat - y) %*% t(w2[-1, , drop = FALSE])
  dw1 <- t(cbind(1, x)) %*% (h * (1 - h) * dh)
  
  w1 <- w1 - learn_rate * dw1
  w2 <- w2 - learn_rate * dw2
  
  list(w1 = w1, w2 = w2)
}

train <- function(x, y, hidden = 5, learn_rate = 1e-2, iterations = 1e4) {
  d <- ncol(x) + 1
  w1 <- matrix(rnorm(d * hidden), d, hidden)
  w2 <- as.matrix(rnorm(hidden + 1))
  for (i in 1:iterations) {
    ff <- feedforward(x, w1, w2)
    bp <- backpropagate(x, y,
                        y_hat = ff$output,
                        w1, w2,
                        h = ff$h,
                        learn_rate = learn_rate)
    w1 <- bp$w1; w2 <- bp$w2
  }
  list(output = ff$output, w1 = w1, w2 = w2)
}


library(ggplot2)
library("emoGG")
theme_set(theme_classic())

set.seed(42)
hotdogs <- two_spirals(labels = c('not The Beach', 'The Beach'))

ggplot(hotdogs) +
  aes(x1, x2, colour = class) +
  geom_point() +
  labs(x = expression(x[1]),
       y = expression(x[2]))

grid <- expand.grid(x1 = seq(min(hotdogs$x1) - 1,
                             max(hotdogs$x1) + 1,
                             by = .1),
                    x2 = seq(min(hotdogs$x2) - 1,
                             max(hotdogs$x2) + 1,
                             by = .1))
grid$class <- factor((predict(logreg, newdata = grid) > 0) * 1,
                     labels = c('not The Beach', 'The Beach'))

x <- data.matrix(hotdogs[, c('x1', 'x2')])
y <- hotdogs$class == 'The Beach'

nodes <- c(1,2,5,10,15,20,24,26,28,30,32,34,36,38)

for(i in seq_along(nodes)){
  message(paste0(i, " of ", length(nodes), ". Nodes = ", nodes[i]))
  
  nnet_i <- train(x, y, hidden = nodes[i], iterations = 1e5)
  
  ff_grid_i <- feedforward(x = data.matrix(grid[, c('x1', 'x2')]),
                           w1 = nnet_i$w1,
                           w2 = nnet_i$w2)
  grid$class <- factor((ff_grid_i$output > .5) * 1,
                       labels = levels(hotdogs$class))
  
  g <- ggplot() +
    geom_raster(data = grid, aes(x1, x2, fill = as.factor(class))) +
    scale_fill_manual(values = c("lightskyblue1", "tan"), name = "Beach?") +
    geom_emoji(data = hotdogs[hotdogs$class == "not The Beach", ], emoji = "1f41f",
               aes(x1, x2)) +
    geom_emoji(data = hotdogs[hotdogs$class == "The Beach", ], emoji = "1f334",
               aes(x1, x2)) +
    coord_equal() +
    theme_void() +
    theme(plot.title = element_text(hjust = 0.5, vjust=-60),
          text=element_text(size=10,  family="Trebuchet MS"),
          legend.position = c(1.1, 0.5)
          
  ggsave(g, file = file.path("path....", paste0(i,".png")), width = 6.5, height = 5)
}