o H&i~@sddlZddlZddlZddlmZddlmZddlmZm Z dgZ ddZ gdZ gd Z gd Zgd Ze egZe egZdd d ZejjddZGdddeZdS)N) constraints) Distribution) broadcast_all lazy_propertyVonMisescCs,t|}|}|r|||}|s |SN)listpop)yZcoefresultr HC:\wamp64\www\opt\env\Lib\site-packages\torch/distributions/von_mises.py _eval_poly s r)g?g$ @g03@g,?N?g2t?gIx?gtHZr?) e3E?g-5?gՒ+Hub?gJNYgTPÂ?g'gZ?gUL+ߐg;^p?)?gY?g(z?g*O?gZ9?g.h?gӰ٩=5?) rg.kg?VmgtZOZ?g<Q g'8`?gP⥝gqJ:N?g;PJ4qcCs|dks |dks J|d}||}t|t|}|dkr#||}|}d|}|d|t|t|}t|dk||}|S)zX Returns ``log(I_order(x))`` for ``x > 0``, where `order` is either 0 or 1. rg@r)r _COEF_SMALLabslog _COEF_LARGEtorchwhere)xorderr ZsmallZlarger r r r _log_modified_bessel_fnAs "rc Cstj|jtj|jd}|sltjd|j|j|jd}|\}}}t t j |} d|| || } ||| } | d| |dk| | d| dkB} | rht| |d| |}|| B}|r|t j |dt j t j S)Ndtypedevice)rrr)rZzerosshapeboolrallZrandrZunbindcosmathpiranyrsignacos) loc concentrationZ proposal_rrdoneuu1u2u3zfcacceptr r r _rejection_sampleXs , r4cseZdZdZejejdZejZdZ dfdd Z ddZ e d d Z e d d Ze d dZeefddZfddZeddZeddZe ddZZS)raX A circular von Mises distribution. This implementation uses polar coordinates. The ``loc`` and ``value`` args can be any real number (to facilitate unconstrained optimization), but are interpreted as angles modulo 2 pi. Example:: >>> # xdoctest: +IGNORE_WANT("non-deterministic") >>> m = VonMises(torch.tensor([1.0]), torch.tensor([1.0])) >>> m.sample() # von Mises distributed with loc=1 and concentration=1 tensor([1.9777]) :param torch.Tensor loc: an angle in radians. :param torch.Tensor concentration: concentration parameter )r)r*FNcs6t||\|_|_|jj}t}t|||dSr)rr)r*r rSizesuper__init__)selfr)r* validate_args batch_shapeZ event_shape __class__r r r7~szVonMises.__init__cCsL|jr|||jt||j}|tdtjt |jdd}|S)Nrrr) _validate_argsZ_validate_sampler*rr#r)r$rr%r)r8valuelog_probr r r r@s  zVonMises.log_probcC|jtjSr)r)tordoubler8r r r _locz VonMises._loccCrAr)r*rBrrCrDr r r _concentrationrFzVonMises._concentrationcCsh|j}ddd|d}|d|d|}d|dd|}d||}t|dk||S)Nrrgh㈵>)rGsqrtrr)r8kappataurho _proposal_rZ_proposal_r_taylorr r r rMs  zVonMises._proposal_rcCs@||}tj||jj|jjd}t|j|j|j | |jjS)a The sampling algorithm for the von Mises distribution is based on the following paper: D.J. Best and N.I. Fisher, "Efficient simulation of the von Mises distribution." Applied Statistics (1979): 152-157. Sampling is always done in double precision internally to avoid a hang in _rejection_sample() for small values of the concentration, which starts to happen for single precision around 1e-4 (see issue #88443). r) Z_extended_shaperemptyrErr)rr4rGrMrB)r8Z sample_shaper rr r r samples  zVonMises.samplecsXzt|WSty+|jd}|j|}|j|}t||||dYSw)Nr>)r9)r6expandNotImplementedError__dict__getr)r*type)r8r:r9r)r*r;r r rPs    zVonMises.expandcC|jS)z8 The provided mean is the circular one. r)rDr r r meansz VonMises.meancCrUrrVrDr r r modesz VonMises.modecCs$dt|jddt|jddS)z< The provided variance is the circular one. rr=r)rr*exprDr r r variances  zVonMises.variancer)__name__ __module__ __qualname____doc__rrealZpositiveZarg_constraintsZsupportZ has_rsampler7r@rrErGrMrZno_gradr5rOrPpropertyrWrXrZ __classcell__r r r;r rhs,       )r)r$rZ torch.jitZtorch.distributionsrZ torch.distributions.distributionrZtorch.distributions.utilsrr__all__rZ_I0_COEF_SMALLZ_I0_COEF_LARGEZ_I1_COEF_SMALLZ_I1_COEF_LARGErrrZjitZscript_if_tracingr4rr r r r s$