Assume you have a source alphabet that includes the $5$ symbols $,($eof$)$ function arithcoding

format long

N$=4$;

$\%$this input sequence represents the example from the screencast

$\%$ "CABBAB#$"$

data$=[3,1,2,2,1,2,4]$

$\%$p$=[\frac{1}{4},\frac{1}{4},\frac{1}{4},\frac{1}{4}]$;

p$=[[\frac{2}{7},\frac{3}{7},\frac{1}{7},\frac{1}{7}]$;

M$=$length$($data$)$;

$\%$encoding

$\%$initialization

upbound $=$ cumsum$($p$)$;

lobound $=[0$ upbound$(1$:end$-1)]$;

for k$1=1$:M

if k$1==$M

finalinterval$=[$lobound$($data$($k$1))$ upbound$($data$($k$1))]$;

else

width $=$ upbound$($data$($k$1))-$ lobound$($data$($k$1))$;

newbottom$=$lobound$($data$($k$1))$;

upbound $=$ newbottom $+$ cumsum$($p$)*$width;

lobound $=[$newbottom upbound$(1$:end$-1)]$;

end

end

finalinterval

$\%$pick any number within the final interval for the decimal code.

$\%$note that picking the optimal binary number is tricky, and depends on

$\%$decimal to binary conversion details.

$\%$we will just cheat and pic the midpoint of the interval

arith codeword $=$ sum$($finalinterval$)/2\%$midpoint of interval

$\%$decoding

upbound $=$ cumsum$($p$)$;

lobound $=[0$ upbound$(1$:end$-1)]$;

decode$=[]$;

finished$=0$;

while finished$==0$

tu $=($arith codeword