B d@sddlZddlmZmZmZmZmZmZmZm Z m Z m Z ddl m Z ddlmZddZddZd d Zdd d ZdddZdddZdddZdddZdS)N) log1plog logaddexpdigitizezerosexp concatenateallcumsumones)uniform)reducecCs:t||ks$tdt|t|f|tt|| S)Nz*cannot take log of negative number %s - %s)r AssertionErrorstrrr)xyrb/work/yifan.wang/ringdown/master-ringdown-env/lib/python3.7/site-packages/lalinference/nest2pos.py logsubexps$rcCsJt|dd|ddtd}t|dd|dd}t||S)zT Trapezoidal integration of given log(func) Returns log of the integral N)rrrr )Zlog_funcZ log_supportZ log_func_sumZlog_dxsrrrlog_integrate_log_trap s"rc Cs>ttdg|d| f}|| d}t|jd}ttt|dtd|td|}tt|}tj |d<|dtd||d<x:t t|dD]&}||td||||d<qWt|||dgf}t||f} t || } t | dd| dd} |dd| dd}|| 8}| |fS)zReturns log_ev, log_wts for the log-likelihood samples in data, assumed to be a result of nested sampling with Nlive live points.z-infNrrgr) rfloatrshaper r lenrnpZNINFrangerr) dataNliveZ start_dataend_datalog_wtsZlog_vols_startZ log_vols_endiZ log_likesZlog_volslog_evZlog_dXsrrrcompute_weightss , &  r$FcsDt|}||}dd|Dttfddtt}||S)zDraw points from the given data (of shape (Nsamples, Ndim)) with associated log(weight) (of shape (Nsamples,)). Draws uniquely so there are no repeated samplescSsg|]}|ttkqSr)rr ).0nrrr :sz"draw_posterior..cs|S)Nr)r") selectionrr;z draw_posterior..)maxlistfilterrr)rr!verboseZmaxWtZnormalised_wtsidxr)r(rdraw_posterior4s r0csxddl}tddt|D\}}|r8tdt|tt|}t|fdd|D|rptdtfddttD}t|fd d|D|rtd td dt||D}|rtd td d|D|rtdtddt|Dg} x8t|D]*\} } t t| d| k} | | | q$W|j | } |rttdt| | S)aDraw samples from the posteriors represented by the (Nruns, Nsamples, Nparams)-shaped array datas, each sampled with the corresponding Nlive number of live points. Will draw without repetition, and weight according to the evidence in each input runrNcSsg|]\}}t|d|qS)logL)r$)r%rrrrrr'Esz'draw_posterior_many..zComputed log_evidences: %scsg|]}t|qSr)r)r%r#)log_max_evidencerrr'Ksz#Relative weights of input files: %scs g|]}|t|qSr)r)r%r")datasfracsrrr'Mscsg|] }|qSrr)r%r&)Ntotrrr'OszDRelative weights of input files taking into account their length: %scSsg|]\}}}t||qSr)r0)r%rlogwtlogZrrrr'RszNumber of input samples: %scSsg|] }t|qSr)r)r%rrrrr'Ssz2Expected number of samples from each input file %scSs g|]\}}t|t|qSr)intr)r%fprrrr'Ts)sizez$Total number of samples produced: %i) astropyzipprintrr rr+rrr appendtablevstack)r3Nlivesr.r<log_evsr!log_total_evidenceNspostsbigpospostfracmaskresultr)r5r3r4r2rdraw_posterior_many>s6   rLc Cs||dkr gSt|jdd}td |d<x*t|D]\}}t|||||d<q6Wtt|d}t||}||dS)zo Draw N samples from the input data, weighted by log_wt. For large N there may be repeated samples rrinf)r;)rrr enumeraterrr r) rr!Nr.Z log_cumsumsr"Zlog_wtusZidxsrrrdraw_N_posterior]s rQc sdddl}tddt||D\}}tt|fdd|D}ddt|||D}|j|S)a Draw Npost samples from the posteriors represented by the (Nruns, Nsamples, Nparams)-shaped array datas, each sampled with the corresponding number of live points Nlive. The returned number of samples may not be exactly Npost due to rounding rNcSsg|]\}}t|d|qS)r1)r$)r%rrrrrr'vsz)draw_N_posterior_many..cs$g|]}ttt|qSr)r8roundr)r%r#)NpostrDrrr'yscSsg|]\}}}t|||qSr)rQ)r%rr6rOrrrr'zs)r<r=r rr@rA) r3rBrSr.r<rCr!rErFr)rSrDrdraw_N_posterior_manyms  rTc s$ddl}tddt|D\}}|r8tdt|tt|}t|fdd|D|rptdtfddttD}t|fd d|D|rtd tg}d dt||D} |rtd td d|D|rtdtddt| D|j |S)abDraw samples from the posteriors represented by the (Nruns, Nsamples, Nparams)-shaped array datas from ROQ runs, each sampled with the corresponding Nlive number of live points. First rescales evidence in each mass prior bin then will draw without repetition, and weight according to the evidence in each input run ala draw_posterior_manyrNcSsg|]\}}t|d|qS)r1)r$)r%rrrrrr'sz0draw_posterior_many_ROQ_runs..zComputed log_evidences: %scsg|]}t|qSr)r)r%r#)r2rrr'sz#Relative weights of input files: %scs g|]}|t|qSr)r)r%r")r3r4rrr'scsg|] }|qSrr)r%r&)r5rrr'szDRelative weights of input files taking into account their length: %scSsg|]\}}}t||qSr)r0)r%rr6r7rrrr'szNumber of input samples: %scSsg|] }t|qSr)r)r%rrrrr'sz2Expected number of samples from each input file %scSs g|]\}}t|t|qSr)r8r)r%r9r:rrrr's) r<r=r>rr rr+rrr@rA) r3rBr.r<rCr!rDrErGrFr)r5r3r4r2rdraw_posterior_many_ROQ_runs}s*  rU)F)F)F)F)F)numpyrrrrrrrrr r r Z numpy.randomr functoolsr rrr$r0rLrQrTrUrrrrs0