def probability(self, condition, **kwargs): cond_inv = False if isinstance(condition, Ne): condition = Eq(condition.args[0], condition.args[1]) cond_inv = True expr = condition.lhs - condition.rhs rvs = random_symbols(expr) z = Dummy('z', real=True, Finite=True) dens = self.compute_density(expr) if any([pspace(rv).is_Continuous for rv in rvs]): from sympy.stats.crv import (ContinuousDistributionHandmade, SingleContinuousPSpace) if expr in self.values: # Marginalize all other random symbols out of the density randomsymbols = tuple(set(self.values) - frozenset([expr])) symbols = tuple(rs.symbol for rs in randomsymbols) pdf = self.domain.integrate(self.pdf, symbols, **kwargs) return Lambda(expr.symbol, pdf) dens = ContinuousDistributionHandmade(dens) space = SingleContinuousPSpace(z, dens) result = space.probability(condition.__class__(space.value, 0)) else: from sympy.stats.drv import (DiscreteDistributionHandmade, SingleDiscretePSpace) dens = DiscreteDistributionHandmade(dens) space = SingleDiscretePSpace(z, dens) result = space.probability(condition.__class__(space.value, 0)) return result if not cond_inv else S.One - result
def __new__(cls, sym, dist): sym = sympify(sym) if isinstance(dist, SingleContinuousDistribution): return SingleContinuousPSpace(sym, dist) if isinstance(dist, SingleDiscreteDistribution): return SingleDiscretePSpace(sym, dist) return Basic.__new__(cls, sym, dist)
def __new__(cls, sym, dist): if isinstance(dist, SingleContinuousDistribution): return SingleContinuousPSpace(sym, dist) if isinstance(dist, SingleDiscreteDistribution): return SingleDiscretePSpace(sym, dist) sym = _symbol_converter(sym) return Basic.__new__(cls, sym, dist)
def pdf(self, *x): dist = self.args[0] z = Dummy('z') if isinstance(dist, ContinuousDistribution): rv = SingleContinuousPSpace(z, dist).value elif isinstance(dist, DiscreteDistribution): rv = SingleDiscretePSpace(z, dist).value return MarginalDistribution(self, (rv,)).pdf(*x)
def __new__(cls, sym, dist): if isinstance(dist, SingleContinuousDistribution): return SingleContinuousPSpace(sym, dist) if isinstance(dist, SingleDiscreteDistribution): return SingleDiscretePSpace(sym, dist) if isinstance(sym, str): sym = Symbol(sym) if not isinstance(sym, Symbol): raise TypeError("s should have been string or Symbol") return Basic.__new__(cls, sym, dist)
def __new__(cls, s, distribution): s = _symbol_converter(s) if isinstance(distribution, ContinuousDistribution): return SingleContinuousPSpace(s, distribution) if isinstance(distribution, DiscreteDistribution): return SingleDiscretePSpace(s, distribution) if isinstance(distribution, SingleFiniteDistribution): return SingleFinitePSpace(s, distribution) if not isinstance(distribution, CompoundDistribution): raise ValueError("%s should be an isinstance of " "CompoundDistribution"%(distribution)) return Basic.__new__(cls, s, distribution)
def test_ContinuousDistributionHandmade(): x = Symbol('x') z = Dummy('z') dens = Lambda(x, Piecewise((S.Half, (0<=x)&(x<1)), (0, (x>=1)&(x<2)), (S.Half, (x>=2)&(x<3)), (0, True))) dens = ContinuousDistributionHandmade(dens, set=Interval(0, 3)) space = SingleContinuousPSpace(z, dens) assert dens.pdf == Lambda(x, Piecewise((1/2, (x >= 0) & (x < 1)), (0, (x >= 1) & (x < 2)), (1/2, (x >= 2) & (x < 3)), (0, True))) assert median(space.value) == Interval(1, 2) assert E(space.value) == Rational(3, 2) assert variance(space.value) == Rational(13, 12)
def _transform_pspace(self, sym, dist, pdf): """ This function returns the new pspace of the distribution using handmade Distributions and their corresponding pspace. """ pdf = Lambda(sym, pdf(sym)) _set = dist.set if isinstance(dist, ContinuousDistribution): return SingleContinuousPSpace(sym, ContinuousDistributionHandmade(pdf, _set)) elif isinstance(dist, DiscreteDistribution): return SingleDiscretePSpace(sym, DiscreteDistributionHandmade(pdf, _set)) elif isinstance(dist, SingleFiniteDistribution): dens = dict((k, pdf(k)) for k in _set) return SingleFinitePSpace(sym, FiniteDistributionHandmade(dens))