zjddlmZddlZddlmZddlZddlmZddlm Z ddl m Z ddl m Z er ddlmZdd lmZGd d e jZdddZdS)) annotationsN) TYPE_CHECKING)check_variable_and_dtype) LayerHelper)exponential_family)in_dynamic_or_pir_mode)Sequence)TensorceZdZUdZded<dfd ZeddZeddZgfdd Z dd Z ddZ ddZ eddZ ddZxZS) Dirichletaa Dirichlet distribution with parameter "concentration". The Dirichlet distribution is defined over the `(k-1)-simplex` using a positive, length-k vector concentration(`k > 1`). The Dirichlet is identically the Beta distribution when `k = 2`. For independent and identically distributed continuous random variable :math:`\boldsymbol X \in R_k` , and support :math:`\boldsymbol X \in (0,1), ||\boldsymbol X|| = 1` , The probability density function (pdf) is .. math:: f(\boldsymbol X; \boldsymbol \alpha) = \frac{1}{B(\boldsymbol \alpha)} \prod_{i=1}^{k}x_i^{\alpha_i-1} where :math:`\boldsymbol \alpha = {\alpha_1,...,\alpha_k}, k \ge 2` is parameter, the normalizing constant is the multivariate beta function. .. math:: B(\boldsymbol \alpha) = \frac{\prod_{i=1}^{k} \Gamma(\alpha_i)}{\Gamma(\alpha_0)} :math:`\alpha_0=\sum_{i=1}^{k} \alpha_i` is the sum of parameters, :math:`\Gamma(\alpha)` is gamma function. Args: concentration (Tensor): "Concentration" parameter of dirichlet distribution, also called :math:`\alpha`. When it's over one dimension, the last axis denotes the parameter of distribution, ``event_shape=concentration.shape[-1:]`` , axes other than last are consider batch dimensions with ``batch_shape=concentration.shape[:-1]`` . Examples: .. code-block:: python >>> import paddle >>> dirichlet = paddle.distribution.Dirichlet(paddle.to_tensor([1., 2., 3.])) >>> print(dirichlet.entropy()) Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True, -1.24434423) >>> print(dirichlet.prob(paddle.to_tensor([.3, .5, .6]))) Tensor(shape=[], dtype=float32, place=Place(cpu), stop_gradient=True, 10.80000019) r concentrationreturnNonec|dkstj|jdkrt d||_t |jdd|jdddS)Nrz:`concentration` parameter must be at least one dimensional)dimmathprodshape ValueErrorr super__init__)selfr __class__s m/lsinfo/ai/hellotax_ai/data_center/backend/venv/lib/python3.11/site-packages/paddle/distribution/dirichlet.pyrzDirichlet.__init__Ss      " "di 0C&D&D&I&IL + ,SbS1=3Frss3KLLLLLcJ|j|jddz S)zbMean of Dirichlet distribution. Returns: Mean value of distribution. rTkeepdim)r sumrs rmeanzDirichlet.mean]s(!D$6$:$:2t$:$L$LLLrc|jdd}|j||jz z|d|dzzz S)zjVariance of Dirichlet distribution. Returns: Variance value of distribution. rTrr)r r!pow)rconcentration0s rvariancezDirichlet.variancefsW+//D/AA"nt7I&IJ   q ! !^a%7 8  rr Sequence[int]ct|tr|nt|}t|j||S)zSample from dirichlet distribution. Args: shape (Sequence[int], optional): Sample shape. Defaults to empty list. ) isinstancetuple _dirichletr expand _extend_shape)rrs rsamplezDirichlet.samplersO $E511CuU||$,33D4F4Fu4M4MNNOOOrvaluecPtj||S)zProbability density function(PDF) evaluated at value. Args: value (Tensor): Value to be evaluated. Returns: PDF evaluated at value. )paddleexplog_probrr1s rprobzDirichlet.prob{s z$--..///rctj||jdz zdtj|jdztj|jdz S)zoLog of probability density function. Args: value (Tensor): Value to be evaluated. ?r)r3logr r!lgammar6s rr5zDirichlet.log_probswZ  $"4s": ; @ @ D DmD.2226677 8mD.//33B77 8 rc|jd}|jjd}tj|jdtj|z ||z tj|zz |jdz tj|jzdz S)zbEntropy of Dirichlet distribution. Returns: Entropy of distribution. rr9)r r!rr3r;digamma)rr'ks rentropyzDirichlet.entropys +//33   $R ( M$, - - 1 1" 5 5mN++ ,>!V^N%C%CC D#c)V^Drms #""""" <<<<<<000000222222333333((((((H=H=H=H=H="4H=H=H=Vr