def log_add_double(x, y):
"""
Return log(x + y) given log(x) and log(y); see [1].
[1] Digital Filtering Using Logarithmic Arithmetic.
Kingsbury and Rayner, 1970.
"""
if "log_add_d" in get_qy().module.global_variables:
log_add_d = Function.get_named("log_add_d")
else:
@Function.define(float, [float, float])
def log_add_d(x_in, y_in):
s = x_in >= y_in
a = qy.select(s, x_in, y_in)
@qy.if_else(a == -numpy.inf)
def _(then):
if then:
qy.return_(-numpy.inf)
else:
qy.return_(a +
qy.log1p(qy.exp(qy.select(s, y_in, x_in) - a)))
return log_add_d(x, y)
def binomial_log_pdf(k, p, n):
"""
Compute the binomial PMF.
"""
name = "binomial_log_pdf_ddd"
if name in get_qy().module.global_variables:
pdf = Function.get_named(name)
else:
@Function.define(float, [float, float, float])
def binomial_log_pdf_ddd(k, p, n):
from qy.math import ln_choose
@qy.if_(k > n)
def _():
qy.return_(-numpy.inf)
@qy.if_(p == 0.0)
def _():
qy.return_(qy.select(k == 0.0, 0.0, -numpy.inf))
@qy.if_(p == 1.0)
def _():
qy.return_(qy.select(k == n, 0.0, -numpy.inf))
qy.return_(ln_choose(n, k) + k * qy.log(p) + (n - k) * qy.log1p(-p))
return binomial_log_pdf_ddd(k, p, n)
def binomial_log_pdf(k, p, n):
"""
Compute the binomial PMF.
"""
name = "binomial_log_pdf_ddd"
if name in get_qy().module.global_variables:
pdf = Function.get_named(name)
else:
@Function.define(float, [float, float, float])
def binomial_log_pdf_ddd(k, p, n):
from qy.math import ln_choose
@qy.if_(k > n)
def _():
qy.return_(-numpy.inf)
@qy.if_(p == 0.0)
def _():
qy.return_(qy.select(k == 0.0, 0.0, -numpy.inf))
@qy.if_(p == 1.0)
def _():
qy.return_(qy.select(k == n, 0.0, -numpy.inf))
qy.return_(
ln_choose(n, k) + k * qy.log(p) + (n - k) * qy.log1p(-p))
return binomial_log_pdf_ddd(k, p, n)
def _ll(self, parameter, sample, out):
"""
Compute log probability under this distribution.
"""
p = parameter.data.load()
k = sample.data.gep(0, 0).load()
n = sample.data.gep(0, 1).load()
if get_qy().test_for_nan:
qy.assert_(p >= 0.0, "invalid p = %s", p)
qy.assert_(p <= 1.0, "invalid p = %s", p)
qy.assert_(k >= 0, "invalid k = %s", k)
qy.assert_(n >= 0, "invalid n = %s", n)
qy.assert_(k <= n, "invalid k = %s (> n = %s)", k, n)
binomial_log_pdf(k, p, n).store(out)
def _ll(self, parameter, sample, out):
"""
Compute log probability under this distribution.
"""
p = parameter.data.load()
k = sample.data.gep(0, 0).load()
n = sample.data.gep(0, 1).load()
if get_qy().test_for_nan:
qy.assert_(p >= 0.0, "invalid p = %s" , p )
qy.assert_(p <= 1.0, "invalid p = %s" , p )
qy.assert_(k >= 0 , "invalid k = %s" , k )
qy.assert_(n >= 0 , "invalid n = %s" , n )
qy.assert_(k <= n , "invalid k = %s (> n = %s)", k, n)
binomial_log_pdf(k, p, n).store(out)
def binomial_pdf(k, p, n):
"""
Compute the binomial PDF function.
"""
name = "gsl_ran_binomial_pdf"
if name in get_qy().module.global_variables:
pdf = Function.get_named(name)
else:
import llvm.core
from ctypes import c_uint
pdf = Function.named(name, float, [c_uint, float, c_uint])
pdf._value.add_attribute(llvm.core.ATTR_READONLY)
pdf._value.add_attribute(llvm.core.ATTR_NO_UNWIND)
return pdf(k, p, n)
def binomial_pdf(k, p, n):
"""
Compute the binomial PDF function.
"""
name = "gsl_ran_binomial_pdf"
if name in get_qy().module.global_variables:
pdf = Function.get_named(name)
else:
import llvm.core
from ctypes import c_uint
pdf = Function.named(name, float, [c_uint, float, c_uint])
pdf._value.add_attribute(llvm.core.ATTR_READONLY)
pdf._value.add_attribute(llvm.core.ATTR_NO_UNWIND)
return pdf(k, p, n)
def log_add_double(x, y):
"""
Return log(x + y) given log(x) and log(y); see [1].
[1] Digital Filtering Using Logarithmic Arithmetic.
Kingsbury and Rayner, 1970.
"""
if "log_add_d" in get_qy().module.global_variables:
log_add_d = Function.get_named("log_add_d")
else:
@Function.define(float, [float, float])
def log_add_d(x_in, y_in):
s = x_in >= y_in
a = qy.select(s, x_in, y_in)
@qy.if_else(a == -numpy.inf)
def _(then):
if then:
qy.return_(-numpy.inf)
else:
qy.return_(a + qy.log1p(qy.exp(qy.select(s, y_in, x_in) - a)))
return log_add_d(x, y)