From 837c7497c40ffed910084b69b67d7dd608e823ac Mon Sep 17 00:00:00 2001
From: "rhys.green" <rhys.green@ligo.org>
Date: Fri, 4 Jan 2019 12:20:52 +0000
Subject: [PATCH] adding custom proposals feature
---
bilby/core/sampler/ptmcmc.py | 84 +++++++++++++++++++++++++-----------
1 file changed, 60 insertions(+), 24 deletions(-)
diff --git a/bilby/core/sampler/ptmcmc.py b/bilby/core/sampler/ptmcmc.py
index 5b20362c0..2ec592771 100644
--- a/bilby/core/sampler/ptmcmc.py
+++ b/bilby/core/sampler/ptmcmc.py
@@ -31,15 +31,16 @@ class PTMCMCSampler(MCMCSampler):
verbose - Update current run-status to the screen (default=False)
"""
- default_kwargs = {'p0' : None , 'Niter' : 10**4, 'ladder' : None
- ,'Tmin' :1 , 'Tmax' : None, 'Tskip' : 100 , 'isave' : 1000
- ,'NUTSweight' : 20 , 'HMCweight' : 20 , 'MALAweight':0
- ,'burn':10000 , 'HMCstepsize' :0.1 , 'HMCsteps':300
- ,'neff' : 10**4 , 'burn' :10**4 , 'thin':1
- ,'covUpdate' : 500, 'SCAMweight':20, 'AMweight':20, 'DEweight':50
- , 'cov': np.eye(1)
- , 'loglargs' : {} , 'loglkwargs' : {} , 'logpargs' : {}, 'logpkwargs': {}
- , 'logl_grad' : None , 'logp_grad' : None, 'outDir' : './ptmcmc_test' , 'verbose': False}
+ default_kwargs = {'p0' : None , 'Niter' : 5 *10**4 ,'neff' : 10**4
+ ,'burn' :10**4 , 'verbose': False
+ ,'ladder' : None,'Tmin' :1 , 'Tmax' : None, 'Tskip' : 100
+ ,'isave' : 1000 ,'thin':1,'covUpdate' : 500
+ ,'SCAMweight':20, 'AMweight':20 , 'DEweight':20
+ ,'HMCweight' : 0 ,'MALAweight':0, 'NUTSweight' : 0
+ ,'HMCstepsize' :0.1 , 'HMCsteps':300
+ ,'groups':None , 'custom_proposals': None
+ ,'loglargs': {} , 'loglkwargs': {} , 'logpargs': {}, 'logpkwargs': {}
+ ,'logl_grad': None , 'logp_grad': None, 'outDir': './ptmcmc_test' }
def __init__(self, likelihood, priors, outdir='outdir', label='label', use_ratio=False, plot=False,
skip_import_verification=False, pos0=None, nburn=None, burn_in_fraction=0.25, **kwargs):
@@ -79,15 +80,17 @@ class PTMCMCSampler(MCMCSampler):
# if 'nsteps' in kwargs:
# kwargs['iterations'] = kwargs.pop('nsteps')
-
+ @property
+ def custom_proposals(self):
+ return self.kwargs['custom_proposals']
@property
def sampler_init_kwargs(self):
keys = [
+ 'groups',
'loglargs',
'logp_grad',
'logpkwargs',
- 'cov',
'loglkwargs',
'logl_grad',
'logpargs',
@@ -116,7 +119,8 @@ class PTMCMCSampler(MCMCSampler):
'Tskip',
'HMCsteps',
'Tmax',
- 'DEweight']
+ 'DEweight'
+ ]
sampler_kwargs = {key: self.kwargs[key] for key in keys}
return sampler_kwargs
@@ -124,38 +128,70 @@ class PTMCMCSampler(MCMCSampler):
def nsteps(self):
return self.kwargs['Niter']
+ @property
+ def nburn(self):
+ return self.kwargs['burn']
+
@nsteps.setter
def nsteps(self, nsteps):
self.kwargs['Niter'] = nsteps
+ @nburn.setter
+ def nburn(self, nsteps):
+ self.kwargs['burn'] = nburn
+
@staticmethod
def _import_external_sampler():
from PTMCMCSampler import PTMCMCSampler
+ import glob
+ import os
+ # OPTIMIZE:
#import acor
#from mpi4py import MPI
#return MPI, PTMCMCSampler
- return PTMCMCSampler
+ return PTMCMCSampler, glob , os
def run_sampler(self):
#MPI , PTMCMCSampler = self._import_external_sampler()
- PTMCMCSampler = self._import_external_sampler()
- #tqdm = get_progress_bar()
- #sampler = emcee.EnsembleSampler(dim=self.ndim, lnpostfn=self.lnpostfn, **self.sampler_init_kwargs)
+ PTMCMCSampler, glob , os = self._import_external_sampler()
init_kwargs = self.sampler_init_kwargs
sampler_kwargs = self.sampler_function_kwargs
sampler = PTMCMCSampler.PTSampler(ndim=self.ndim, logp = self.log_prior,
- logl = self.log_likelihood, **init_kwargs)
+ logl = self.log_likelihood, cov= np.eye(self.ndim),
+ **init_kwargs)
tqdm = get_progress_bar()
- print(self.nsteps)
- sampler.sample(p0 = self.p0 , **sampler_kwargs)
-
+ if self.custom_proposals is not None :
+ for proposal in self.custom_proposals:
+ print('adding ' + str(proposal) + ' to proposals with weight:'
+ + str(self.custom_proposals[proposal][1]))
+ sampler.addProposalToCycle(self.custom_proposals[proposal][0]
+ ,self.custom_proposals[proposal][1])
+ else :
+ pass
+ sampler.sample(p0 = self.p0 , **sampler_kwargs)
+ #### The next bit is very hacky, the ptmcmc writes the samples and
+ #### other info to file so here i read this info, write it to the result
+ #### object then delete it
+ data =np.loadtxt('ptmcmc_test/chain_1.txt')
+ #jumpfiles = glob.glob('ptmcmc_test/*jump.txt')
+ #jumps = map(np.loadtxt, jumpfiles)
+ samples = data[:,:-4]
+ log_post = data[:, -4]
+ loglike = data[:,-3]
+ acceptance_rate = data[:,-2]
+ pt_swap_accept = data[:,-1]
+ for f in glob.glob('./ptmcmc_test/*'):
+ os.remove(f)
+ os.rmdir('ptmcmc_test')
self.result.sampler_output = np.nan
- self.calculate_autocorrelation(sampler.chain.reshape((-1, self.ndim)))
- self.print_nburn_logging_info()
+ #self.calculate_autocorrelation(sampler.chain.reshape((-1, self.ndim)))
+ #self.print_nburn_logging_info()
self.result.nburn = self.nburn
- self.result.samples = sampler.chain[:, self.nburn:, :].reshape((-1, self.ndim))
- self.result.walkers = sampler.chain
+ self.result.samples = samples[self.nburn:]
+ #### Walkers isn't really applicable here but appears to be needed to
+ #### turn samples into data frame
+ self.result.walkers = samples[self.nburn:]
self.result.log_evidence = np.nan
self.result.log_evidence_err = np.nan
return self.result
--
GitLab