from functools import singledispatch from sympy.core.symbol import Dummy from sympy.functions.elementary.exponential import exp from sympy.utilities.lambdify import lambdify from sympy.external import import_module from sympy.stats import DiscreteDistributionHandmade from sympy.stats.crv import SingleContinuousDistribution from sympy.stats.crv_types import ChiSquaredDistribution, ExponentialDistribution, GammaDistribution, \ LogNormalDistribution, NormalDistribution, ParetoDistribution, UniformDistribution, BetaDistribution, \ StudentTDistribution, CauchyDistribution from sympy.stats.drv_types import GeometricDistribution, LogarithmicDistribution, NegativeBinomialDistribution, \ PoissonDistribution, SkellamDistribution, YuleSimonDistribution, ZetaDistribution from sympy.stats.frv import SingleFiniteDistribution scipy = import_module("scipy", import_kwargs={'fromlist':['stats']}) @singledispatch def do_sample_scipy(dist, size, seed): return None # CRV @do_sample_scipy.register(SingleContinuousDistribution) def _(dist: SingleContinuousDistribution, size, seed): # if we don't need to make a handmade pdf, we won't import scipy.stats z = Dummy('z') handmade_pdf = lambdify(z, dist.pdf(z), ['numpy', 'scipy']) class scipy_pdf(scipy.stats.rv_continuous): def _pdf(dist, x): return handmade_pdf(x) scipy_rv = scipy_pdf(a=float(dist.set._inf), b=float(dist.set._sup), name='scipy_pdf') return scipy_rv.rvs(size=size, random_state=seed) @do_sample_scipy.register(ChiSquaredDistribution) def _(dist: ChiSquaredDistribution, size, seed): # same parametrisation return scipy.stats.chi2.rvs(df=float(dist.k), size=size, random_state=seed) @do_sample_scipy.register(ExponentialDistribution) def _(dist: ExponentialDistribution, size, seed): # https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.expon.html#scipy.stats.expon return scipy.stats.expon.rvs(scale=1 / float(dist.rate), size=size, random_state=seed) @do_sample_scipy.register(GammaDistribution) def _(dist: GammaDistribution, size, seed): # https://stackoverflow.com/questions/42150965/how-to-plot-gamma-distribution-with-alpha-and-beta-parameters-in-python return scipy.stats.gamma.rvs(a=float(dist.k), scale=float(dist.theta), size=size, random_state=seed) @do_sample_scipy.register(LogNormalDistribution) def _(dist: LogNormalDistribution, size, seed): # https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.lognorm.html return scipy.stats.lognorm.rvs(scale=float(exp(dist.mean)), s=float(dist.std), size=size, random_state=seed) @do_sample_scipy.register(NormalDistribution) def _(dist: NormalDistribution, size, seed): return scipy.stats.norm.rvs(loc=float(dist.mean), scale=float(dist.std), size=size, random_state=seed) @do_sample_scipy.register(ParetoDistribution) def _(dist: ParetoDistribution, size, seed): # https://stackoverflow.com/questions/42260519/defining-pareto-distribution-in-python-scipy return scipy.stats.pareto.rvs(b=float(dist.alpha), scale=float(dist.xm), size=size, random_state=seed) @do_sample_scipy.register(StudentTDistribution) def _(dist: StudentTDistribution, size, seed): return scipy.stats.t.rvs(df=float(dist.nu), size=size, random_state=seed) @do_sample_scipy.register(UniformDistribution) def _(dist: UniformDistribution, size, seed): # https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.uniform.html return scipy.stats.uniform.rvs(loc=float(dist.left), scale=float(dist.right - dist.left), size=size, random_state=seed) @do_sample_scipy.register(BetaDistribution) def _(dist: BetaDistribution, size, seed): # same parametrisation return scipy.stats.beta.rvs(a=float(dist.alpha), b=float(dist.beta), size=size, random_state=seed) @do_sample_scipy.register(CauchyDistribution) def _(dist: CauchyDistribution, size, seed): return scipy.stats.cauchy.rvs(loc=float(dist.x0), scale=float(dist.gamma), size=size, random_state=seed) # DRV: @do_sample_scipy.register(DiscreteDistributionHandmade) def _(dist: DiscreteDistributionHandmade, size, seed): from scipy.stats import rv_discrete z = Dummy('z') handmade_pmf = lambdify(z, dist.pdf(z), ['numpy', 'scipy']) class scipy_pmf(rv_discrete): def _pmf(dist, x): return handmade_pmf(x) scipy_rv = scipy_pmf(a=float(dist.set._inf), b=float(dist.set._sup), name='scipy_pmf') return scipy_rv.rvs(size=size, random_state=seed) @do_sample_scipy.register(GeometricDistribution) def _(dist: GeometricDistribution, size, seed): return scipy.stats.geom.rvs(p=float(dist.p), size=size, random_state=seed) @do_sample_scipy.register(LogarithmicDistribution) def _(dist: LogarithmicDistribution, size, seed): return scipy.stats.logser.rvs(p=float(dist.p), size=size, random_state=seed) @do_sample_scipy.register(NegativeBinomialDistribution) def _(dist: NegativeBinomialDistribution, size, seed): return scipy.stats.nbinom.rvs(n=float(dist.r), p=float(dist.p), size=size, random_state=seed) @do_sample_scipy.register(PoissonDistribution) def _(dist: PoissonDistribution, size, seed): return scipy.stats.poisson.rvs(mu=float(dist.lamda), size=size, random_state=seed) @do_sample_scipy.register(SkellamDistribution) def _(dist: SkellamDistribution, size, seed): return scipy.stats.skellam.rvs(mu1=float(dist.mu1), mu2=float(dist.mu2), size=size, random_state=seed) @do_sample_scipy.register(YuleSimonDistribution) def _(dist: YuleSimonDistribution, size, seed): return scipy.stats.yulesimon.rvs(alpha=float(dist.rho), size=size, random_state=seed) @do_sample_scipy.register(ZetaDistribution) def _(dist: ZetaDistribution, size, seed): return scipy.stats.zipf.rvs(a=float(dist.s), size=size, random_state=seed) # FRV: @do_sample_scipy.register(SingleFiniteDistribution) def _(dist: SingleFiniteDistribution, size, seed): # scipy can handle with custom distributions from scipy.stats import rv_discrete density_ = dist.dict x, y = [], [] for k, v in density_.items(): x.append(int(k)) y.append(float(v)) scipy_rv = rv_discrete(name='scipy_rv', values=(x, y)) return scipy_rv.rvs(size=size, random_state=seed)