Off all thinkable numbers, Pi has somewhat of celebrity status. It is so famous it even has its holiday. Mathematicians who use the American way to write dates recognise March the 14th as Pi Day. Some authors have even assigned a mystical status to the number of Pi. In the novel Contact by Carl Sagan, the heroine Ellie discovers a hidden message in the base-11 representation of Pi. In the 1998 film Pi by Darren Aronofsky, the protagonist is driven mad by the idea that a secret pattern in random numbers. The string of numbers that form the decimals of Pi might seem random, but they are of course perfectly predictable and calculable. Their apparent randomness can create artful visualisations.

This article discusses some approximations of Pi using the R language and visualises the results. The video below is an ode to the aesthetic beauty of the digits of Pi by Numberphile.

Approximations of Pi

My favourite approximation of Pi has served me well through the early part of my engineering career. When using calculators that did not have a button for , I used 22/7. It is only accurate to two decimals, but that is more than enough for dredging projects.

The base R language can provide 22 decimals of Pi using options(digits = 22). If you need more accuracy, then the MPFR package in R provides a library for multiple-precision floating-point computations. The Const function can compute Pi to a high level of precision with thousands of digits.

library(rmpfr)
pi_mpfr <- Const("pi", prec = 200000)

There are also more creative ways to determine the number Pi using Monte Carlo simulations. Imagine you are terribly bored, sitting inside a room with floorboards. You decide to take a match with length  and the floorboards are wide. The needle is half the length of the width of the floorboards (). You drop the matchstick randomly for about ten million times () and you record every time the needle crosses the groove of one of the boards (). The number Pi can be approximated by:

This bit of code simulates dropping the needle ten million times, which gives an estimated value of . You would have to be extremely bored to throw a needle that many times to only achieve ccuracy to two decimals!

library(tidyverse)
t <- 2
l <- 1
n <- 10000000

needles <- data_frame(phi = runif(n, 0, 2 * pi),
           x1 = runif(n, 0, (l + 3)), 
           y1 = runif(n, 0, (l + 3)),
           x2 = x1 + l * sin(phi),
           y2 = y1 + l * cos(phi), 
           overlap = (x1 < 1 & x2 > 1) | (x1 < 3 & x2 > 3))

ggplot(needles[1:1000,], aes(x1, y1)) + 
  geom_segment(aes(xend = x2, yend = y2, colour = overlap)) + 
  geom_vline(xintercept = c(1, 3), colour = "red") +
  scale_color_manual(values = c("gray", "black")) + 
  theme_void() 
ggsave("Misc/buffon.png", dpi = 300)

pi_buffon <- (n * l) / (sum(needles$overlap) * t)

Visualising the number Pi

In 1888, the John Venn, inventor of the eponymous Venn diagram, wanted to visualise the apparent randomness of the digits of Pi. He obtained the first 707 digits of Pi from amateur mathematician William Shanks. Unfortunately, only the first 527 decimals of Pi were correct but this was only discovered in 1944. Venn assigned a compass point to the digits 0 to 7 and then drew lines to show the path indicated by each digit, ignoring the digits 8 and 9.

Running through the values of Pi this way produces a line that snakes its way through the graph. When you use random numbers with the same method, the graph looks very similar. In this case, I have downloaded the digits of Pi from the On-Line Encyclopedia of Integer Sequences (Nr. A000796).

library(tidyverse)
# Download Pi Digits
pi_digits <- read.csv("http://oeis.org/A000796/b000796.txt", header = FALSE, sep = " ", skip = 1) %>%
  select(digit = V2)

# Venn walk
venn_walk <- function(digits) {
  digits <- digits[digits != 8 & digits != 9]
  l <- length(digits) - 1
  x <- rep(0, l)
  y <- x
  for (i in 1:l) {
    a <- digits[i + 1] * pi / 4
    dx <- round(sin(a))
    dy <- round(cos(a))
    x[i + 1] <- x[i] + dx 
    y[i + 1] <- y[i] + dy
  }
  coords <- data_frame(x = x, y = y) 
  ggplot(coords, aes(x, y)) + geom_path() + 
    geom_point(data = coords[c(1, l + 1), ], aes(x, y), colour = "red", size = 2) + 
    theme_void() 
} 
venn_walk(pi_digits) 
ggsave("Misc/venn_pi_walk.png", dpi = 300) 

# Random Numbers
data.frame(digit = sample(0:7, 20000, replace = TRUE)) %>%
 venn_walk()

The Numberphile video shows some beautiful visualisations of Pi, one of which I like to share with you to close this article. Martin Krzywinski created this, and many other, visualisations of Pi.

data_frame(x = rep(1:20, 20),
           y = unlist(lapply(1:20, function(x) rep(x, 20))),
           d = pi_digits$digit[1:400]) %>%
  mutate(d = factor(d)) %>%
  ggplot(aes(x, y, colour = d)) + geom_point(size = 3) +
  theme_void() + 
  theme(plot.background = element_rect(fill = "black"),
        panel.background = element_rect(fill = "black"),
        legend.position="none")
ggsave("Misc/pi_dots.png")

As always, you can find the latest version of this code on GitHub. Feel free to subscribe to this blog if you like to receive articles in your mailbox.

The post Approximations of Pi: A Random Walk though the Beauty of Pi appeared first on The Devil is in the Data.