gam::rnd Namespace Reference

Random number functions. More...

Functions
template<class T >
T &	cond (T &v, const T &va, const T &vb, float pab, float pba)
	Conditional probability with two values.
template<class T >
T	geom (int n, T mul, T start=1, float p=0.5)
	Returns ith element in geometric series.
template<class T >
T	neg (T val, float prob=0.5f)
	Returns value negated with probability, otherwise the value.
template<class T >
const T &	pick (const T &val1, const T &val2, float prob=0.5f)
	Returns val1 with probability, otherwise val2.
template<class T >
bool	prob (T &rng, float p=0.5f)
	Returns true with a probability of p.
bool	prob (float p=0.5f)
	Returns true with a probability of p.
bool	prob (char c)
	Characters 0-8 return true with probability c/8.
void	push (uint32_t seed=0)
	Push current RNG state onto stack (stack size = 1).
void	pop ()
	Pop RNG state from stack.
template<class T >
void	permute (T *arr, uint32_t len)
	Randomly permutes (shuffles) elements in array.
float	quan (uint32_t q)
	Returns value in [0, 1) quantized by q divisions.
template<class T >
T	quanOct (uint32_t q, T o)
	Returns value in [o, 2*o) quantized by q divisions.
void	seed (uint32_t value=0)
	Seed global PRNG. If seed is 0, then the system time (in seconds) is used.
template<class T >
void	set (T *arr, uint32_t len, uint32_t num, T val=1)
	Randomly set a certain amount of elements to a value.
template<class T >
void	thin (T *arr, uint32_t len, float prob=0.5f)
	Zeroes elements according to a probability.
template<class T >
T	uniExc (const T &exc, const T &max, const T &min=T(0))
	Returns uniform random within interval [min, max) excluding 'exc' argument.
template<class T >
uint32_t	weighted (T *weights, uint32_t num, T weightsSum=(T) 1)
	Returns random integer in [0, num) according to weights (a PDF).

---

Detailed Description

Random number functions.

Unless specified, these operations use an internal Tausworthe RNG, favoring quality over a slight deficiency in speed compared to a linear congruential RNG.

The following template functions are available:

 // Returns random number linearly mapped to [bound1, bound2)
 T distr(T bound2, T bound1 = 0);

 // Fills array with random numbers in [0, 1)
 void distr(T * dst, uint32_t len);

 // Fills array with random numbers linearly mapped to [bound1, bound2)
 void distr(T * dst, uint32_t len, T bound2, T bound1 = 0);

 // Copies random elements from 'src' to 'dst'.
 void distr(T * dst, uint32_t len, T * src, uint32_t srcLen);
  *

where "distr" is one of the following random number distributions:

add2 - sum of 2 uniform values (triangle)
add2I - inverse pdf of add2
add3 - sum of 3 uniform values (2nd order normal curve)
binS - uniform signed binary
lin - linear decrease
mul2 - product of 2 uniform values
pow2 - uniform value squared
pow3 - uniform value cubed
uni - uniform value

Since casts between floats and generic types are used in the ranged varieties, signed integer bounds will always be exclusive.

---

Function Documentation

T & cond	(	T &	v,
		const T &	va,
		const T &	vb,
		float	pab,
		float	pba
	)

Conditional probability with two values.

'pab' and 'pba' are the probabilities of a given b and b given a.

bool prob

(

char

c

)

Characters 0-8 return true with probability c/8.

In addition to numeric characters, '/' returns true and '.' returns false.

void push

(

uint32_t

seed = 0

)

Push current RNG state onto stack (stack size = 1).

After pushing, the current RNG is seeded with 'seed' unless 'seed' = 0.

uint32_t weighted	(	T *	weights,
		uint32_t	num,
		T	weightsSum = (T)1
	)

Returns random integer in [0, num) according to weights (a PDF).

If the weights are not normalized, then the proper weightsSum must be passed in.