def _get_weighted_statistics(self,data,weights):
n, tot = super(NegativeBinomialIntegerR,self)._get_weighted_statistics(data,weights)
if n > 0:
alpha_n, betas_n = self.alpha_0 + tot, self.beta_0 + self.r_support*n
data, weights = flattendata(data), flattendata(weights)
log_marg_likelihoods = \
special.betaln(alpha_n, betas_n) \
- special.betaln(self.alpha_0, self.beta_0) \
+ (special.gammaln(data[:,na]+self.r_support)
- special.gammaln(data[:,na]+1) \
- special.gammaln(self.r_support)).dot(weights)
else:
log_marg_likelihoods = np.zeros_like(self.r_support)
return n, tot, log_marg_likelihoods
def _get_weighted_statistics(self,data,weights):
n, tot = super(NegativeBinomialIntegerR,self)._get_weighted_statistics(data,weights)
if n > 0:
alpha_n, betas_n = self.alpha_0 + tot, self.beta_0 + self.r_support*n
data, weights = flattendata(data), flattendata(weights)
log_marg_likelihoods = \
special.betaln(alpha_n, betas_n) \
- special.betaln(self.alpha_0, self.beta_0) \
+ (special.gammaln(data[:,na]+self.r_support)
- special.gammaln(data[:,na]+1) \
- special.gammaln(self.r_support)).dot(weights)
else:
log_marg_likelihoods = np.zeros_like(self.r_support)
return n, tot, log_marg_likelihoods
def _get_statistics(self,data):
n = getdatasize(data)
if n > 0:
data = flattendata(data)
feasible = self.r_support <= data.min()
assert np.any(feasible)
r_support = self.r_support[feasible]
normalizers = (special.gammaln(data[:,na]) - special.gammaln(data[:,na]-r_support+1)
- special.gammaln(r_support)).sum(0)
return n, data.sum(), normalizers, feasible
else:
return n, None, None, None
def _get_statistics(self,data):
n = getdatasize(data)
if n > 0:
data = flattendata(data)
feasible = self.r_support <= data.min()
assert np.any(feasible)
r_support = self.r_support[feasible]
normalizers = (special.gammaln(data[:,na]) - special.gammaln(data[:,na]-r_support+1)
- special.gammaln(r_support)).sum(0)
return n, data.sum(), normalizers, feasible
else:
return n, None, None, None
def _get_statistics(self,data=[]):
n, tot = self._fixedr_distns[0]._get_statistics(data)
if n > 0:
data = flattendata(data)
alphas_n, betas_n = self.alphas_0 + tot, self.betas_0 + self.r_support*n
log_marg_likelihoods = \
special.betaln(alphas_n, betas_n) \
- special.betaln(self.alphas_0, self.betas_0) \
+ (special.gammaln(data[:,na]+self.r_support)
- special.gammaln(data[:,na]+1) \
- special.gammaln(self.r_support)).sum(0)
else:
log_marg_likelihoods = np.zeros_like(self.r_support)
return log_marg_likelihoods
def resample(self,data=[],niter=20):
if getdatasize(data) == 0:
self.p = np.random.beta(self.alpha_0,self.beta_0)
self.r = np.random.gamma(self.k_0,self.theta_0)
else:
data = atleast_2d(flattendata(data))
N = len(data)
for itr in range(niter):
### resample r
msum = sample_crp_tablecounts(self.r,data).sum()
self.r = np.random.gamma(self.k_0 + msum, 1/(1/self.theta_0 - N*np.log(1-self.p)))
### resample p
self.p = np.random.beta(self.alpha_0 + data.sum(), self.beta_0 + N*self.r)
return self
def _get_statistics(self,data=[]):
n, tot = self._fixedr_distns[0]._get_statistics(data)
if n > 0:
data = flattendata(data)
alphas_n, betas_n = self.alphas_0 + tot, self.betas_0 + self.r_support*n
log_marg_likelihoods = \
special.betaln(alphas_n, betas_n) \
- special.betaln(self.alphas_0, self.betas_0) \
+ (special.gammaln(data[:,na]+self.r_support)
- special.gammaln(data[:,na]+1) \
- special.gammaln(self.r_support)).sum(0)
else:
log_marg_likelihoods = np.zeros_like(self.r_support)
return log_marg_likelihoods
def resample(self,data=[],niter=20):
if getdatasize(data) == 0:
self.p = np.random.beta(self.alpha_0,self.beta_0)
self.r = np.random.gamma(self.k_0,self.theta_0)
else:
data = np.atleast_2d(flattendata(data))
ones = np.ones(data.shape[1],dtype=float)
for itr in range(niter):
### resample r
msum = sample_crp_tablecounts(float(self.r),data,ones).sum()
self.r = np.random.gamma(self.k_0 + msum, 1/(1/self.theta_0 - N*np.log(1-self.p)))
### resample p
self.p = np.random.beta(self.alpha_0 + data.sum(), self.beta_0 + N*self.r)
return self
def resample_python(self,data=[],niter=20):
if getdatasize(data) == 0:
self.p = np.random.beta(self.alpha_0,self.beta_0)
self.r = np.random.gamma(self.k_0,self.theta_0)
else:
data = flattendata(data)
N = len(data)
for itr in range(niter):
### resample r
msum = 0.
for n in data:
msum += (np.random.rand(n) < self.r/(np.arange(n)+self.r)).sum()
self.r = np.random.gamma(self.k_0 + msum, 1/(1/self.theta_0 - N*np.log(1-self.p)))
### resample p
self.p = np.random.beta(self.alpha_0 + data.sum(), self.beta_0 + N*self.r)
return self
def resample_python(self,data=[],niter=20):
if getdatasize(data) == 0:
self.p = np.random.beta(self.alpha_0,self.beta_0)
self.r = np.random.gamma(self.k_0,self.theta_0)
else:
data = flattendata(data)
N = len(data)
for itr in range(niter):
### resample r
msum = 0.
for n in data:
msum += (np.random.rand(n) < self.r/(np.arange(n)+self.r)).sum()
self.r = np.random.gamma(self.k_0 + msum, 1/(1/self.theta_0 - N*np.log(1-self.p)))
### resample p
self.p = np.random.beta(self.alpha_0 + data.sum(), self.beta_0 + N*self.r)
return self
def resample_logseriesaug(self,data=[],niter=20):
# an alternative algorithm, kind of opaque and no advantages...
if getdatasize(data) == 0:
self.p = np.random.beta(self.alpha_0,self.beta_0)
self.r = np.random.gamma(self.k_0,self.theta_0)
else:
data = flattendata(data)
N = data.shape[0]
logF = self.logF
L_i = np.zeros(N)
data_nz = data[data > 0]
for itr in range(niter):
logR = np.arange(1,logF.shape[1]+1)*np.log(self.r) + logF
L_i[data > 0] = sample_discrete_from_log(logR[data_nz-1,:data_nz.max()],axis=1)+1
self.r = np.random.gamma(self.k_0 + L_i.sum(), 1/(1/self.theta_0 - np.log(1-self.p)*N))
self.p = np.random.beta(self.alpha_0 + data.sum(), self.beta_0 + N*self.r)
return self
def resample_logseriesaug(self,data=[],niter=20):
# an alternative algorithm, kind of opaque and no advantages...
if getdatasize(data) == 0:
self.p = np.random.beta(self.alpha_0,self.beta_0)
self.r = np.random.gamma(self.k_0,self.theta_0)
else:
data = flattendata(data)
N = data.shape[0]
logF = self.logF
L_i = np.zeros(N)
data_nz = data[data > 0]
for itr in range(niter):
logR = np.arange(1,logF.shape[1]+1)*np.log(self.r) + logF
L_i[data > 0] = sample_discrete_from_log(logR[data_nz-1,:data_nz.max()],axis=1)+1
self.r = np.random.gamma(self.k_0 + L_i.sum(), 1/(1/self.theta_0 - np.log(1-self.p)*N))
self.p = np.random.beta(self.alpha_0 + data.sum(), self.beta_0 + N*self.r)
return self
def _get_statistics(self,data):
# NOTE: since this isn't really in exponential family, this method needs
# to look at hyperparameters. form posterior hyperparameters for the p
# parameters here so we can integrate them out and get the r statistics
n, tot = super(NegativeBinomialIntegerR,self)._get_statistics(data)
if n > 0:
alpha_n, betas_n = self.alpha_0 + tot, self.beta_0 + self.r_support*n
data = flattendata(data)
log_marg_likelihoods = \
special.betaln(alpha_n, betas_n) \
- special.betaln(self.alpha_0, self.beta_0) \
+ (special.gammaln(data[:,na]+self.r_support)
- special.gammaln(data[:,na]+1) \
- special.gammaln(self.r_support)).sum(0)
else:
log_marg_likelihoods = np.zeros_like(self.r_support)
return n, tot, log_marg_likelihoods
def _get_statistics(self,data):
# NOTE: since this isn't really in exponential family, this method needs
# to look at hyperparameters. form posterior hyperparameters for the p
# parameters here so we can integrate them out and get the r statistics
n, tot = super(NegativeBinomialIntegerR,self)._get_statistics(data)
if n > 0:
alpha_n, betas_n = self.alpha_0 + tot, self.beta_0 + self.r_support*n
data = flattendata(data)
log_marg_likelihoods = \
special.betaln(alpha_n, betas_n) \
- special.betaln(self.alpha_0, self.beta_0) \
+ (special.gammaln(data[:,na]+self.r_support)
- special.gammaln(data[:,na]+1) \
- special.gammaln(self.r_support)).sum(0)
else:
log_marg_likelihoods = np.zeros_like(self.r_support)
return n, tot, log_marg_likelihoods