\;j%/ddlZddlmZddlZddlZddlmZmZddl m Z ddl m Z ddl mZddlmZGdd e j"Zy) N)Iterable) check_type convert_dtype)Variable) distribution)in_dynamic_mode)randomcleZdZdZd fd ZedZedZd dZd dZ dZ dZ d Z d Z xZS)Normala The Normal distribution with location `loc` and `scale` parameters. Mathematical details The probability density function (pdf) is .. math:: pdf(x; \mu, \sigma) = \frac{1}{Z}e^{\frac {-0.5 (x - \mu)^2} {\sigma^2} } .. math:: Z = (2 \pi \sigma^2)^{0.5} In the above equation: * :math:`loc = \mu`: is the mean. * :math:`scale = \sigma`: is the std. * :math:`Z`: is the normalization constant. Args: loc(int|float|list|tuple|numpy.ndarray|Tensor): The mean of normal distribution.The data type is float32 and float64. scale(int|float|list|tuple|numpy.ndarray|Tensor): The std of normal distribution.The data type is float32 and float64. name(str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`. Examples: .. code-block:: python >>> import paddle >>> from paddle.distribution import Normal >>> # Define a single scalar Normal distribution. >>> dist = Normal(loc=0., scale=3.) >>> # Define a batch of two scalar valued Normals. >>> # The first has mean 1 and standard deviation 11, the second 2 and 22. >>> dist = Normal(loc=[1., 2.], scale=[11., 22.]) >>> # Get 3 samples, returning a 3 x 2 tensor. >>> dist.sample([3]) >>> # Define a batch of two scalar valued Normals. >>> # Both have mean 1, but different standard deviations. >>> dist = Normal(loc=1., scale=[11., 22.]) >>> # Complete example >>> value_tensor = paddle.to_tensor([0.8], dtype="float32") >>> normal_a = Normal([0.], [1.]) >>> normal_b = Normal([0.5], [2.]) >>> sample = normal_a.sample([2]) >>> # a random tensor created by normal distribution with shape: [2, 1] >>> entropy = normal_a.entropy() >>> print(entropy) Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True, [1.41893852]) >>> lp = normal_a.log_prob(value_tensor) >>> print(lp) Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True, [-1.23893857]) >>> p = normal_a.probs(value_tensor) >>> print(p) Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True, [0.28969154]) >>> kl = normal_a.kl_divergence(normal_b) >>> print(kl) Tensor(shape=[1], dtype=float32, place=Place(cpu), stop_gradient=True, [0.34939718]) c tslt|dtttj t ttfdt|dtttj t ttfdd|_ ||nd|_ d|_ t|tr t|}t|tr t|}|j||r*||_||_t!|j|_ nWt|trt|trd|_ t|tj r)t#|jdvr|j|_ nBt|tj r(t#|jdvr|j|_ |j%||\|_|_|jt!|jjk7r`t'j(|j|j|_t'j(|j|j|_t*|Y|jj.y) Nlocr scaleFfloat32T)rfloat64dtype)rrintfloatnpndarrayrlisttupleall_arg_is_floatnamer isinstance_validate_argsr rrstr _to_tensorpaddlecastsuper__init__shape)selfr rr __class__s cG:\00. PROJECTS\API\Inventory\templateJSON\kerjaOCR\Lib\site-packages\paddle/distribution/normal.pyr"zNormal.__init__as eRZZ4?   eRZZ4?  !& ,D(  c3 *C eS !%LE   sE *DHDJ&syy1DJ#u%*UE*B(,%#rzz*s399~B0!YY E2::.3u{{3CH4#[[ #'??3#> DHdjzz]488>>::!;;txxtzzB#[[4::F  (c|jS)z[Mean of multinomial distribuion. Returns: Tensor: mean value. )r r$s r&meanz Normal.means xxr'c8|jjdS)zbVariance of lognormal distribution. Returns: Tensor: variance value. )rpowr)s r&variancezNormal.varianceszz~~a  r'c6t|ts tdtst |dt dt |}t |j|jzj}|jdz}d|vr||z}t ||z}tj|j|jzdj|d<tj|d|j}tj||}tj|} t!j"| dd||j } | ||jzz} tj$| |j| } | S||z}t!j"|dd||j tj&||j |jzz} tj$| |j| } |j(rtj| || S| S) a&Generate samples of the specified shape. Args: shape (Sequence[int], optional): Shape of the generated samples. seed (int): Python integer number. Returns: Tensor, A tensor with prepended dimensions shape.The data type is float32. %sample shape must be Iterable object.seedsample_sampler?)r*stdr1rrr)rr TypeErrorrrrrr rr#rritemfullrreshaper gaussianaddzerosr) r$r#r1 batch_shaper output_shape fill_shapezero_tmpzero_tmp_reshapezero_tmp_shapenormal_random_tmpoutputs r&r2z Normal.samples%*CD D tVcH 5U DHHtzz1889 yy9$   ;.LkE12J"LLDJJ)>?BGGIJqM{{:sDJJ?H%~~h E #\\*:;N &ScDJJ! '*:TZZ*GHFZZtShannon entropy in nats. The entropy is .. math:: entropy(\sigma) = 0.5 \log (2 \pi e \sigma^2) In the above equation: * :math:`scale = \sigma`: is the std. Returns: Tensor, Shannon entropy of normal distribution.The data type is float32. _entropyr4rr5?r,r8) rrr rr#rr:rr;r>mathlogpi)r$rr@rB fill_dtyperCs r&entropyzNormal.entropys"yy:%DHHtzz1889  k*J"LLDJJ)>?BGGIJqM((TZZ/66J{{:sJ?H{{;TZZ@Hzz (N $((1tww;' '&**TZZ(5J*K K  r'c |jdz}|j|j|}|j|jz}t j |j}t j d||jz ||jz zzd|zz |tj tjdtjzz|S)zLog probability density/mass function. Args: value (Tensor): The input tensor. Returns: Tensor: log probability.The data type is same with :attr:`value` . _log_prob@r8) r_check_values_dtype_in_probsr rrrQsubtractrPsqrtrR)r$valuervar log_scales r&log_probzNormal.log_probsyy;&11$((EBjj4::%JJtzz*  UTXX%%$((*:; <c J 3=!9: :  r'c|jdz}|j|j|}|j|jz}t j t j d||jz ||jz zzd|zz tjdtjz|jz|S)zProbability density/mass function. Args: value (Tensor): The input tensor. Returns: Tensor, probability. The data type is same with :attr:`value` . _probsrWrXr,r8) rrYr rrdivideexprPr[rR)r$r\rr]s r&probsz Normal.probssyy8#11$((EBjj4::%}} JJ)edhh.>?@C#IN YYq477{ #djj 0   r'cXtst|dtd|jdz}|j|jz }||z}|j |j z |jz }||z}t jd|zd|dz t j|z z|S)aThe KL-divergence between two normal distributions. The probability density function (pdf) is .. math:: KL\_divergence(\mu_0, \sigma_0; \mu_1, \sigma_1) = 0.5 (ratio^2 + (\frac{diff}{\sigma_1})^2 - 1 - 2 \ln {ratio}) .. math:: ratio = \frac{\sigma_0}{\sigma_1} .. math:: diff = \mu_1 - \mu_0 In the above equation: * :math:`loc = \mu_0`: is the mean of current Normal distribution. * :math:`scale = \sigma_0`: is the std of current Normal distribution. * :math:`loc = \mu_1`: is the mean of other Normal distribution. * :math:`scale = \sigma_1`: is the std of other Normal distribution. * :math:`ratio`: is the ratio of scales. * :math:`diff`: is the difference between means. Args: other (Normal): instance of Normal. Returns: Tensor, kl-divergence between two normal distributions.The data type is float32. other kl_divergence_kl_divergencerOr6r8) rrr rrr rr>rQ)r$rfr var_ratiot1s r&rgzNormal.kl_divergence+sB ugv ?yy++JJ,  ) hh"ekk 1 "Wzz )OSBHvzz)/D$DED  r')N)r)rk)__name__ __module__ __qualname____doc__r"propertyr*r.r2rLrTr_rdrg __classcell__)r%s@r&r r sXBH-)^!!+Z+" @ * ,+ r'r )rPcollections.abcrnumpyrrpaddle.base.data_feederrrpaddle.base.frameworkrpaddle.distributionrpaddle.frameworkr paddle.tensorr Distributionr rkr'r&rzs6 $ =*,, z \ & &z r'