Skip to content

Latest commit

 

History

History
76 lines (56 loc) · 1.78 KB

README.textile

File metadata and controls

76 lines (56 loc) · 1.78 KB

Bayes

Bayes is a Domain-Specific Language for probability theory. The goal is to mimic mathematical notation as closely as possible. There is an emphasis on choice in the API – many expressions can be written in several different ways, allowing you to choose between being concise and explicit. I have grand plans of expanding it into a full probability/statistics package, but for now the project is much more modest.

So, what does it look like?

I’m glad you asked!

Working with boolean random variables:


  p = Bayes::P.new
  a = Bayes::RV.new

p[a] = { true => 0.1, false => 0.9 }
#or…
p[a] = [0.1, 0.9]
#or even…
p[a] = [0.1] #final value is inferred
p[ a[true] ] = 0.1
p[+a] #=> 0.1
p[~a] #=> 0.9
p[a] #=> [0.1, 0.9]

Discrete random variables:


  p = Bayes::P.new
  a = Bayes::RV.new [:yes, :no, :maybe], [0.1, 0.4, 0.5]

p[ a[:yes] ] #=> 0.1
p[ a.maybe ] = 0.7
p[ a.no ] = 0.2
p[a] #=> [0.1, 0.2, 0.7]

TODO

Package as a gem

Compound Variables
Something like this:


  p = Bayes::P.new
  a = Bayes::RV.new :bool, [0.1, 0.9]
  b = Bayes::RV.new :bool

p[b & a] = { :true =>  { :true => 0.25, :false => 0.75 },
:false => { :true => 0.1, :false => 0.9 } }
p[+b & ~a] #=> 0.75

Conditional Variables
Something like this:


  p = Bayes::P.new
  a = Bayes::RV.new :bool, [0.1, 0.9]
  b = Bayes::RV.new :bool

p[b|a] = { :true =>  { :true => 0.25, :false => 0.75 },
:false => { :true => 0.1, :false => 0.9 } }
p[+b|~a] #=> 0.75

Continuous Variables
Something like this:


  p = Bayes::P.new
  a = Bayes::RV.new :continuous, :normal => { :mean => 1, :variance: 0.2 }

p[ a[1.2] ] = #=> whatever this is