Instead of a truly random number, you wish to randomly select a value from a set in which some values are more likely than others. For example, you may wish to simulate a normal distribution (i.e., a "bell curve") for a set of data.
We will give a recipe for generating numbers with a normal distribution (aka Gaussian distribution, the bell shaped one). No existing library supports this functionality, though inspiration for one can be gained by viewing the Schemathics CVS. We will discuss a do-it-yourself method for explanatory purposes.
You will have to determine what kind of distribution you want, and locate the appropriate algorithm from a statistics reference.
For this recipe, we will consider the normal (Gaussian) distribution. If you need other distributions see either the CVS or consult a numerical analyst.
The function dis_var_new
returns a stochastic variable (a thunk) with mean mu
and standard deviance sigma
.:
% derived from example in the documentation of SRFI27 % and translated to Erlang -record( dist_state, {state, mu, sigma} ). dist_var_new(Mu, Sigma) -> dist_var_new(false, Mu, Sigma). % create the thunk dist_var_new(State, Mu, Sigma) -> This = #dist_state { state = State, mu = 1.0 * Mu, sigma = 1.0 * Sigma}, IntPid = spawn(cookbook, dispatch, [This]), fun () -> IntPid ! {self(), value}, receive {retval, Any} -> Any end end. dispatch(This) -> receive {Pid, value} -> {NewThis, Value} = value(This), Pid!{retval, Value}, dispatch(This) end. value(This) -> case This#dist_state.state of true -> Val = This#dist_state.mu + (This#dist_state.sigma * This#dist_state.state), {This#dist_state{state = false}, Val}; _ -> sigma_loop(This) end. sigma_loop(This) -> V1 = 2.0 * random:uniform() - 1.0, V2 = 2.0 * random:uniform() - 1.0, S = (V1 * V1) + (V2 * V2), if S >= 1.0 -> sigma_loop(This); true -> Scale = math:sqrt( (-2.0 * math:log(S)) / S), Val = This#dist_state.mu + (This#dist_state.sigma * Scale * V1), {This#dist_state{state = Scale * V2}, Val} end. |
An example of usage:
1> X=dist_var_new(0, 1). #Fun<cookbook.4.38595032> 2> X1(). -0.873932 3> X1(). 4.91005e-2 4> X1(). 1.55993e-2 5> X1(). 0.181456 |
6> random:seed() |
The algorithm used is the polar Box Muller method. The algorithm takes two independent uniformly distributed random numbers between 0 and 1 (present in the code as random:uniform()
) and generates two numbers with a mean of my and standard deviation sigma. Note that the method produces two numbers at a time. Since we only need one, the second is saved for later in the variable next
.
Note that the Perl Cookbook includes an interesting discussion of converting a set of values (and weights) into a distribution. This should also be converted to Erlang and shown here.
Mathematically-inclined Erlangers should also take a good look at Schemathics, which contains these and many other statistical methods.
-- BrentAFulgham - 14 May 2004 -- JensAxelSoegaard - 01 Jun 2004
[TODO: Move the following remarks to another recipe]
If you wish to randomly select from a set of weights and values, convert the weights into a probability distribution, then use the resulting distribution to pick a value.
If you have a list of weights and values you want to randomly pick from, follow this two-step process: First, turn the weights into a probability distribution with weight_to_dist below, and then use the distribution to randomly pick a value with weighted_rand:
-- BrentAFulgham - 25 Aug 2004
CookbookForm | |
---|---|
TopicType: | Recipe |
ParentTopic: | NumberRecipes |
TopicOrder: | 100 |